Accepted Manuscript Comparative Clinical Effectiveness of Management Strategies for Sciatica: Systematic Review and Network Meta-Analyses Ruth Lewis, Nefyn H. Williams, Alex J. Sutton, Kim Burton, NafeesUd Din, Hosam E. Matar, Maggie Hendry, Ceri J. Phillips, Sadia Nafees, Deborah Fitzsimmons, I. Rickard, Clare Wilkinson PII:
S1529-9430(13)01497-6
DOI:
10.1016/j.spinee.2013.08.049
Reference:
SPINEE 55546
To appear in:
The Spine Journal
Received Date: 10 November 2011 Revised Date:
9 July 2013
Accepted Date: 23 August 2013
Please cite this article as: Lewis R, Williams NH, Sutton AJ, Burton K, Din N, Matar HE, Hendry M, Phillips CJ, Nafees S, Fitzsimmons D, Rickard I, Wilkinson C, Comparative Clinical Effectiveness of Management Strategies for Sciatica: Systematic Review and Network Meta-Analyses, The Spine Journal (2013), doi: 10.1016/j.spinee.2013.08.049. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
COMPARATIVE CLINICAL EFFECTIVENESS OF MANAGEMENT STRATEGIES FOR SCIATICA: SYSTEMATIC REVIEW AND NETWORK META-ANALYSES Ruth Lewis1, Nefyn H Williams1,2, Alex J Sutton3, Kim Burton4, NafeesUd Din1, Hosam E Matar5, Maggie Hendry1, Ceri J Phillips6, Sadia Nafees1, Deborah Fitzsimmons4, I Rickard7,
1
RI PT
Clare Wilkinson1
North Wales Centre for Primary Care Research, College of Health & Behavioural Sciences,
Bangor University, Wrexham, UK
North Wales Organisation for Randomised Trials in Health (NWORTH), Bangor University,
SC
2
Bangor, UK
Department of Health Sciences, University of Leicester, Leicester, UK
4
Spinal Research Institute, University of Huddersfield, Huddersfield, UK
M AN U
3
5
Sheffield Teaching Hospitals NHS Foundation Trust, Northern General Hospital, Sheffield, UK School of Human and Health Sciences, Swansea University, Swansea, UK
7
Betws-y-Coed, UK
Ruth Lewis - Lecturer
TE D
6
Nefyn H Williams – Senior clinical lecturer Alex J Sutton – Professor
EP
Kim Burton - Professor
NafeesUd Din – Research Associate
AC C
Hosam E Matar – Foundation II doctor Maggie Hendry - Research Fellow Ceri J Phillips - Professor SadiaNafees – Research Assistant Deborah Fitzsimmons - Senior Lecturer Ian Rickard - Patient representative Clare Wilkinson - Professor
Maggie Hendry; Ceri J Phillips; Sadia Nafees; Deborah Fitzsimmons; Ian Rickard; Clare Wilkinson
ACCEPTED MANUSCRIPT
Corresponding author:
RI PT
Ruth Lewis North Wales Centre for Primary Care Research North Wales Clinical School College of Health & Behavioural Sciences Bangor University Gwenfro Units 6-7 Wrexham Technology Park Wrexham LL13 7YP
SC
Tel +44 (0)1978 727340 Fax +44 (0)1978 311419 e-mail: [email protected]
Acknowledgement: We would like to thank Barbara France and Annie Hendry who
M AN U
helped with producing the Figures and Tables; Di Pasterfield for proof reading the manuscript; and Ian Braithwaite (Consultant Surgeon) and Rob Chakraverty (Sports Physician) for providing clinical input at various stages of the review.We would also like to thank the associated editor and the anonymous reviewers for their careful and constructive criticism which helped improve and simplify a previous version of this
TE D
paper.
This study was funded by the National Institute for Health Research (NIHR) HTA programme (project number: 06/79/01) and is published in full in Health Technology Assessment; Vol.15. The funding body had no role in design and conduct of the study; data collection; management, analyses, and interpretation of the data;
EP
preparation of the manuscript; or the decision to submit the article for publication. The views and opinions expressed therein are those of the authors and do not
AC C
necessarily reflect those of the HTA programme, NIHR, NHS or the Department of Health.
ACCEPTED MANUSCRIPT
1
COMPARATIVE CLINICAL EFFECTIVENESS OF MANAGEMENT STRATEGIES FOR
2
SCIATICA: SYSTEMATIC REVIEW AND NETWORK META-ANALYSES
3 4 ABSTRACT
6
Background: There are numerous treatment approaches for sciatica. Previous systematic
7
reviews have not compared all these strategies together.
8
Purpose: To compare the clinical effectiveness of different treatment strategies for sciatica
9
simultaneously.
SC
RI PT
5
Study design: Systematic review and network meta-analysis.
11
Methods: We searched 28 electronic databases and online trial registries, along with
12
bibliographies of previous reviews for comparative studies evaluating any intervention to
13
treat sciatica in adults, with outcome data on global effect or pain intensity. Network meta-
14
analysis methods were used to simultaneously compare all treatment strategies and allow
15
indirect comparisons of treatments between studies. The study was funded by the UK
16
National Institute for Health Research (NIHR) HTA programme; there are nopotential conflict
17
of interests.
18
Results: We identified 122 relevant studies; 90 were randomised controlled trials (RCTs) or
19
quasi-RCTs. Interventions were grouped into 21 treatment strategies. Internal and external
20
validity of included studies was very low. For overall recovery as the outcome, compared
21
with inactive control or conventional care, there was a statistically significant improvement
22
following disc surgery, epidural injections, non-opioid analgesia, manipulation, and
23
acupuncture. Traction, percutaneous discectomy and exercise therapy were significantly
24
inferior to epidural injections or surgery. For pain as the outcome, epidural injections, and
25
biological agents were significantly better than inactive control, but similar findings for disc
26
surgery were not statistically significant. Biological agents were significantly better for pain
27
reduction than bed rest, non-opioids, and opioids. Opioids, education/advice alone, bed rest,
AC C
EP
TE D
M AN U
10
1
ACCEPTED MANUSCRIPT
and percutaneous discectomy were inferior to most other treatment strategies; although
2
these findings represented large effects, they were statistically equivocal.
3
Conclusions: For the first time many different treatment strategies for sciatica have been
4
compared in the same systematic review and meta-analysis. This approach has provided
5
new data to assist shared decision-making. The findings support the effectiveness of non-
6
opioid medication, epidural injections and disc surgery. They also suggest that spinal
7
manipulation, acupuncture, and experimental treatments such as anti-inflammatory
8
biological agents, may be considered. The findings do not provide support for the
9
effectiveness of opioid analgesia, bed rest, exercise therapy, education/advice (when used
SC
RI PT
1
alone), percutaneous discectomy or traction. The issue of how best to estimate the
11
effectiveness of treatment approaches according to their order within a sequential treatment
12
pathway remains an important challenge.
13 14
M AN U
10
INTRODUCTION
16
Sciatica is the term used for the syndrome characterised by radicular leg pain, with or
17
without sensory deficits, radiating along the distribution of the sciatic nerve.1-3 In about 90%
18
of cases, it is caused by an intervertebral disc herniation resulting in nerve root irritation.4-6 It
19
is a common reason for seeking medical advice,7,8 and has considerable economic
20
consequence in terms of healthcare resources and lost productivity.7 The diagnosis and
21
management of sciatica varies considerably within and between countries,4 which may
22
reflect treatment availability, clinician preference and socio-economic variables rather than
23
evidence-based practice.
EP
AC C
24
TE D
15
25
Previous systematic reviews (including meta-analyses) have evaluated the effectiveness of
26
various individual treatment approaches for sciatica, including conservative treatments,9-12
27
epidural steroid injections,9,11,13,14 and surgical procedures.15 However, numerous treatments
28
have not been directly compared. Furthermore, in order to choose the optimal treatment(s), it
2
ACCEPTED MANUSCRIPT
would be more helpful if all candidate treatments could be compared in the same analysis,
2
as opposed to using a series of simple but inefficient standard pairwise meta-analyses
3
comparing only two treatments at a time. It has been acknowledged that there is difficulty in
4
interpreting the findings of multiple comparisons with low power, due to the small number of
5
participants or events, which are inclined to result in statistically insignificant findings.16,17
RI PT
1
6
A network meta-analysis,18 by contrast, enables the simultaneous comparison of more than
8
two treatment approaches, whilst combining data derived from both direct within-study
9
comparisons between two treatment strategies (e.g A vs B) and comparisons constructed
10
from two studies that have one treatment in common (e.g. A vs B, B vs C).17 This type of
11
analysis can only be applied to connected networks of randomised controlled trials (RCTs),19
12
but preserves the within-trial randomised comparison of each study19 and allows information
13
on treatment strategies to be “borrowed” from other studies within the network, thereby
14
increasing the total sample size.20,21 Network meta-analysis conducted using Bayesian
15
methods22-24 also allows the treatment strategies to be ranked in terms of clinical
16
effectiveness with an estimate of the probability that each strategy is ‘best’.25
TE D
M AN U
SC
7
17
Our primary aims were to simultaneously compare the clinical effectiveness of different
19
treatment strategies for sciatica using network meta-analyses, in order to identify the best
20
treatment and to provide estimates for all possible pairwise comparisons, based on both
21
direct and indirect evidence. Our secondary aims were to demonstrate the feasibility of using
22
network meta-analyses as a rational basis for clinical decision making when a number of
23
treatment options are available and where a series of conventional systematic reviews have
24
failed to help with real-world treatment decisions. The analyses presented in this paper
25
represent a refinement of initial network meta-analyses conducted as part of a broader
26
Health Technology Assessment (HTA) evaluating the clinical and cost effectiveness of
27
treatments for sciatica. A full account of the study methods and literature search are
28
presented in the HTA monograph (which also includes the protocol).16
AC C
EP
18
3
ACCEPTED MANUSCRIPT
1 METHODS
3
Search strategy
4
Included studies were identified via an extensive literature search described in full, including
5
the search strategy, in the HTA monograph.16 The search incorporated 28 electronic
6
databases and trial registries including MEDLINE, EMBASE, and AMED. Databases were
7
searched from inception until December 2009 without language restriction. The reference
8
lists of previous systematic reviews and included studies were also scanned for further
9
references.
SC
RI PT
2
M AN U
10 Study selection and data extraction
12
This review included any comparative study (experimental or observational) with adults who
13
had sciatica diagnosed clinically, or where clinical imaging confirmed lumbar disc prolapse
14
consistent with the clinical findings. The essential clinical criterion was radicular leg pain
15
worse than back pain.16 Studies of sciatica caused by conditions other than a prolapsed
16
intervertebral disc were included if it was documented that radicular leg pain was worse than
17
back pain. If imaging was used, it had to demonstrate evidence of nerve root compromise.
18
Studies that included participants with non-specific low back pain were only included if the
19
findings for patients with sciatica were reported separately. Any type of intervention to treat
20
sciatica was considered. These were categorised, for the purpose of the present analyses,
21
into one of 21 categories (See Table 1). Interventions that included a combination of more
22
than one treatment strategy (or mixed treatments) were excluded from the network meta-
23
analyses due to uncertainty regarding the extent of interaction between the combined
24
interventions. The same applied to post-surgical interventions due to surgery being included
25
as a separate treatment category. Studies comparing interventions that were grouped under
26
the same treatment strategy were also excluded. Three further studies evaluating
27
experimental interventions for sciatica (common peroneal nerve block,26 protolytic enzyme,28
28
and colchicine30) were excluded from the analyses as these interventions did not fit the
AC C
EP
TE D
11
4
ACCEPTED MANUSCRIPT
1
treatment categorisation. For the present network meta-analyses we concentrated on overall
2
response and pain intensity rather than back specific function, so three studies which only
3
reported outcome data for back specific function were excluded.32-34
5
RI PT
4 [Table 1: Treatment categorisation]
6
Two reviewers screened studies for inclusion independently. Data were extracted by one
8
reviewer and checked by a second using the original paper, whilst quality assessment was
9
done by two reviewers independently. Any disagreements were resolved by discussion. The
SC
7
quality of both trials and observational studies was assessed using the same checklist,
11
which was based on one used by the Back Review Group of the Cochrane Collaboration for
12
RCTs35 and another recommended by the Guidelines for Systematic Reviews in Health
13
Promotion and Public Health Taskforce36 (developed by the Effective Public Health Practice
14
Project, Canada37). The criteria covered external validity, selection bias and confounding,
15
detection bias, performance bias, and attrition bias. Studies were coded as strong, moderate
16
or weak for each domain, estimating the risk of bias.
17
TE D
M AN U
10
Outcome measures
19
Overall response or global effect was analysed as a binary outcome (treatment success vs
20
failure) and synthesised using odds ratios (ORs). Where studies reported overall response in
21
terms of both overall improvement and improvement in leg pain, the data on overall
22
improvement were used. For studies that reported both physician and patient perceived
23
global effect, the data for patients’ perceived effect were used.
AC C
24
EP
18
25
Pain intensity (on a scale of 0-100) was analysed as a continuous outcome measure using
26
weighted mean difference (WMD). We only included pain assessment from one location
27
from each study using the preference hierarchy of leg pain then overall pain. Where feasible,
28
missing data were estimated from the published data, using standard methods, such as
5
ACCEPTED MANUSCRIPT
standard deviations (SDs) derived from standard errors (SEs).38 Where mean values were
2
unavailable but the medians were reported, these were used instead. If SDs for baseline
3
values were available these were substituted for missing SDs. For studies that did not report
4
sufficient data to derive the SDs, they were imputed using the weighted mean,39 which was
5
calculated separately for each intervention category.
RI PT
1
6 Statistical analysis
8
The network meta-analyses were based on a single time point, using the findings from
9
individual studies closest to six months follow-up. Sensitivity analyses were conducted to
10
assess the impact of excluding non-randomised studies (observational studies and non-
11
RCTs).
M AN U
SC
7
12
The network meta-analyses were conducted using a hierarchical random-effects model18
14
within the Bayesian framework. Bayesian methods are based on the idea that unknown
15
quantities, such as population means or proportions, have probability distributions.23 You
16
start with a distribution that is based on prior knowledge or subjective belief about the
17
population and then update this using data from your included studies. However, using non-
18
informative priors (such as, a normal distribution with a large variance) means that the
19
results are based predominantly on the data from the included studies, and as such will
20
mirror those obtained using frequentist or classical meta-analysis methods. Bayesian
21
methods are implemented using model-based simulations, which means that they can be
22
used to perform complex analyses that incorporate multiple data sources and allow for
23
various parameter uncertainties within a single coherent model, which is why we chose to
24
use these methods.
AC C
EP
TE D
13
25 26
Our network meta-analyses were conducted using WinBUGS1.4.3 software,40 which uses
27
Markov chain Monte Carlo (MCMC) simulation methods to run thousands of simulated
28
iterations based on the data and description of the proposed distributions for relevant
6
ACCEPTED MANUSCRIPT
parameters. The iterative simulations area generally started at multiple points in order to
2
ensure the samples are drawn from the whole sampling frame. The first 50,000 iterations (or
3
burn-in) were discarded, and the results are based on a further sample of at least 100,000
4
simulations, ensuring that the multiple simulation strings have converged and distributions
5
were informed by later simulations. Numerical methods such as the Brooks-Gelman-Rubin
6
statistic40 and the inspection of the auto-correlation and history plots, which are routine
7
assessments made when using MCMC methods, were used to check that convergence had
8
occurred. The model fit was checked by the global goodness of fit statistic, residual
9
deviance. If the model is an adequate fit, it is expected that the residual deviance should be
SC
RI PT
1
roughly equal to the number of data points.19 Non-informative priors were used for normal
11
distributions for means, and uniform distributions for standard deviations. The treatment
12
strategy ‘inactive control’ was used as the reference treatment. This included interventions
13
that represent the non (active) treatment of sciatica, such as no treatment, sham treatment,
14
or placebo (two studies used active placebo). The WinBUGS codes (or models) that we
15
used are presented in Supplementary Material (Web Appendix A). The robustness of the
16
network meta-analyses were also evaluated by comparing the findings (where head to head
17
studies were available) with those of standard ‘direct’ pairwise meta-analyses16 conducted
18
using a random-effects model41 based on frequentist methods22-24 in Stata 10.
TE D
EP
19
M AN U
10
The assumptions of a random-effects network meta-analysis are that (1) the treatment
21
effects are additive (i.e. the relative effect of treatment A vs C can be estimated from the
22
effect of A vs B and B vs C);19,42,43 (2) study-specific treatment effects are drawn from a
23
common distribution (exchangeable);19,44 and (3) this common distribution or heterogeneity is
24
constant between the different comparisons.19,44 We evaluated heterogeneity between
25
studies, defined as the variability of the results across studies within each treatment
26
comparison over and above chance,45 by examining the findings of standard pairwise meta-
27
analyses using visual inspection of the forest plots, as well as Chi2 statistic to test for the I2
28
statistic to quantify statistical heterogeneity.46,47
AC C
20
7
ACCEPTED MANUSCRIPT
1 2 RESULTS
4
Included studies
5
As seen in Figure 1, 122 studies were included in the revised network meta-analyses48-168
6
(one publication included two studies122), 86 were RCTs51,52,54,55,57,59-65,67,69,70,72-74,76-79,81-86,88,90-
7
93,96,98-100,102-104,106,108-110,113,114,117-127,129-131,133,135-144,146-148,152-158,161,162,164,167
8
RCTs.49,94,107,112 The network meta-analysis of global effect included 95 studies (68 RCTs/Q-
9
RCTs) and pain intensity 53 studies (46 RCTs/Q-RCTs). A description of the interventions,
RI PT
3
SC
and four Q-
populations, study design, and outcome data for the pairwise studies are presented in
11
Supplementary Material (Web Appendix B).
M AN U
10
12
Eleven (9%) studies had a strong overall quality rating61,82,100,102,119,131,135,143,155,158,164 and
14
eight (7%) had a strong overall external validity rating;100,103,119,124,131,143,156,158 five (4%) of
15
which had a strong rating for both.100,119,131,143,158 Only 26 (21%) studies used both adequate
16
randomisation and adequate or partially adequate (using sealed envelopes, n=16) allocation
17
concealment.
TE D
13
18
The proportion of studies that limited inclusion to patients with acute sciatica (duration of
20
symptoms 10),
18
but these were not statistically significant: chemonucleolysis (E), traction (H), exercise
19
therapy (K), passive physical therapy (such as ultrasound and transcutaneous electrical
20
nerve stimulation) (L), bed rest (N), opioid medication (O), percutaneous discectomy (Q),
21
intradiscal injections (S), and radio frequency treatment (U), all of which were associated
22
with very wide confidence intervals. This reflects the limited evidence available for biological
23
agents, which included a small placebo controlled RCT (n=24) that reported a large effect
24
estimate in favour of biological agents (OR 10.0; 95% CI: 0.65, 166.67. see Supplementary
25
Material Table C1).
AC C
EP
TE D
M AN U
10
26 27
The results of the sensitivity analyses excluding observational studies and non-RCTs
28
showed broad agreement with the main analyses. For global effect, the most notable
10
ACCEPTED MANUSCRIPT
discrepancies occurred with biological agents compared with chemonucleolysis,
2
conventional, and care exercise therapy. A more detailed narrative of the differences
3
between the analyses with and without the non-randomised studies is presented in the
4
Supplementary Material (Web Appendix D)
RI PT
1
5 Pain intensity
7
In terms of pain intensity, the only treatment comparisons with inactive control that were
8
statistically significant were epidural injections (D) and biological agents (M). Biological
9
agents, which had the highest probability of being best (0.33), were also found to be
SC
6
statistically significantly better at reducing pain than non-opioids (F), bed rest (N), opioids (O)
11
and radio frequency treatment (U); these findings were all associated with wide credible
12
intervals. When considering the magnitude of effect, bed rest (N), education/advice alone
13
(P), percutaneous discectomy (Q), and radiofrequencly treatment (U) tended to fare worse
14
when compared with most treatment strategies, with findings showing a non-statistically
15
significant difference of more than 25 points. Acupuncture (J), had the second highest
16
probability of being best (0.19) and resulted in reductions of pain intensity of more than 25
17
points compared with bed rest, opioids, education/advice alone, percutaneous discectomy
18
and radio frequency treatment, none of which were statistically significant and all had wide
19
credible intervals.
TE D
EP
20
M AN U
10
For pain intensity the most notable discrepancies between the network meta-analysis with
22
and without observational studies and non-RCTs only occurred with biological agents (vs
23
inactive control, conventional care, disc surgery, non-opioids, intra-operative interventions,
24
acupuncture, exercise therapy, opioids, and neuropathic painmodulators). Biological agents
25
no longer had the highest probability of being best (0.03; see Supplementary material Table
26
C4). These discrepancies are likely to be due to the small number of included studies with a
27
limited number of participants evaluating biological agents (2 RCTs n=131; 1 non-
28
randomised RCT n=72; and 1 historical cohort study n=10).
AC C
21
11
ACCEPTED MANUSCRIPT
1 2 Between study heterogeneity, model fit and comparison with standard pairwise meta-
4
analyses
5
Based on the Gelman-Rubin statistic, convergence occurred at around 6-8000 iterations for
6
both outcome measures (global effect, pain intensity). The auto-correlation and history plots
7
also showed good convergence. The goodness of fit of the models to the data, measured by
8
the residual deviance, was found to be good for all three outcomes (Supplementary Material,
9
Web Appendix E).
SC
RI PT
3
M AN U
10
The results of the evaluation of between-study heterogeneity showed a moderate to high
12
level16 of statistical heterogeneity for many of the pairwise comparisons, as well as across all
13
studies as a whole. The heterogeneity was greater for the analysis of pain intensity than
14
global effect, with an I2 statistic of less than 75% (i.e. moderate or less) for all but one
15
pairwise comparison (epidural injections vs conventional care). The observed values for I2
16
are presented in Figure 1. Heterogeneity did not improve when non-randomised studies
17
were removed.
TE D
11
18
The comparison of the results from the network meta-analyses with that of the conventional
20
pairwise meta-analyses showed broad agreement with slightly more discrepancies for the
21
analyses of pain intensity. These discrepancies were greatest for comparisons that had very
22
little direct evidence, such as biological agents.
24
AC C
23
EP
19
25
DISCUSSION
26
This is the first systematic review that has included all treatment strategies for sciatica in the
27
same analysis using a network meta-analysis method that includes indirect comparisons.
28
The advantages of such analyses are that they can simultaneously compare more than two
12
ACCEPTED MANUSCRIPT
treatments in the same coherent analysis; provide relative effect estimates for all treatment
2
comparisons, even those that have not been directly compared in head to head trials; enable
3
the estimation of the probability that each treatment is best; and reduce the uncertainty in the
4
treatment effect estimates.
RI PT
1
5 Summary of results
7
In terms of overall response or global effect, there was a statistically significant improvement
8
following disc surgery, epidural injections, non-opioid medication, intra-operative
9
interventions, manipulation, and acupuncture when compared with inactive control or
SC
6
conventional care. Epidural injections, disc surgery, and intra-operative interventions
11
were also statistically significantly better than percutaneous discectomy, chemonucleolysis,
12
intradiscal injections, and radiofrequencly treatment; with epidural injections, and intra-
13
operative interventions also statistically significantly better than both traction, and exercise
14
therapy. While biological agents and acupuncture had the highest probability of being best
15
and had the largest effect estimates when compared with inactive control, these findings
16
were associated with very wide credible intervals, reflecting the lack of information on these
17
effect estimates.
TE D
M AN U
10
EP
18
In terms of pain intensity, there was a statistically significant reduction in pain following
20
epidural injections and biological agents compared with inactive control, but there was no
21
significant difference between disc surgery and inactive control. Biological agents had the
22
highest probability of being best, and were also statistically significantly better than non-
23
opioid medication, opioid medication, bed rest, and radio frequency treatment. However,
24
when the analysis was restricted to RCTs, biological agents no longer had the highest
25
probability of being best and were not found to be statistically better than any other
26
treatments. When considering the magnitude of effect, bed rest, education/advice alone,
27
percutaneous discectomy, and radiofrequencly treatment were considerably inferior when
AC C
19
13
ACCEPTED MANUSCRIPT
1
compared with most treatment strategies, but these findings were not statistically significant
2
and were associated with wide credible intervals.
3 Overall, the results of the sensitivity analyses excluding non-randomised studies showed
5
broad agreement with the main analyses, with the findings generally becoming non-
6
statistically significant due to broader credible intervals for the analyses restricted to RCTs
7
and Q-RCTs. The most notable discrepancies occurred with treatment strategies that were
8
associated with a small number of included studies such as those reporting treatment with
9
biological agents.
SC
RI PT
4
M AN U
10 Findings of previous reviews
12
Previous reviews of non-surgical treatments have either found no evidence of
13
effectiveness,9,10 conflicting evidence,11,12 or have reached different conclusions concerning
14
the effectiveness of epidural steroid injections.9,11,13,14,169,170 A Cochrane systematic review of
15
surgical interventions did not combine the results of four RCTs comparing discectomy with
16
non-surgical treatment due to heterogeneity, and concluded that the results showed a
17
temporary benefit of disc surgery at one year follow-up.15 In that review the effectiveness of
18
discectomy was justified by using informal indirect comparison of chemonucleolysis with
19
placebo, and chemonucleolysis with disc surgery; chemonucleolysis was more effective than
20
placebo and discectomy more effective than chemonucleolysis, therefore disc surgery was
21
superior to placebo. Using our network meta-analyses, it was possible to make a more
22
robust statement on disc surgery compared with placebo: disc surgery was statistically
23
significantly better than placebo in terms of global effect but not for pain intensity.
24
AC C
EP
TE D
11
25
Strengths and weaknesses
26
One of the main strengths of our network meta-analyses is the wide range of treatment
27
strategies used to treat sciatica that were not only considered in the same review, but
28
compared simultaneously in the same analysis. Another strength is that they were based on
14
ACCEPTED MANUSCRIPT
a systematic and comprehensive search of the literature up (until December 2009) that
2
covered any therapeutic intervention for sciatica. Although we acknowledge that these
3
searches are not current, and as such, more recent relevant data is likely to have been
4
excluded.
RI PT
1
5
The RCT is widely regarded as the design of choice when assessing the effectiveness of
7
health care interventions171 and we acknowledge the controversy over the inclusion of non-
8
randomised evidence. Non-randomised studies were included in the search because some
9
treatment approaches may not have been evaluated by RCTs, and also to increase the
SC
6
precision of the findings for interventions evaluated by a limited number of studies.
11
Observational studies can have better external validity than RCTs172,173 and provide more
12
generalisable findings. However, observational studies are likely to be affected by selection
13
bias and confounding, and may therefore yield estimates of association that deviate from the
14
true underlying relationship beyond the play of chance.174 As it happens, most of the RCTs
15
did not report the method of generating the randomisation sequence or allocation
16
concealment, which means that selection bias or confounding might still be present.
17
Excluding the non-randomised studies in a sensitivity analysis did not affect the structure of
18
the network and the overall findings of both series of network meta-analyses were similar,
19
although less precise for the analyses of RCTs.
TE D
EP
20
M AN U
10
Network meta-analysis methods enabled us to go beyond the pairwise comparisons reported
22
in previous systematic reviews. They allowed us to simultaneously compare all the available
23
treatment strategies for sciatica and provided estimates of relative treatment effects for all
24
conceivable comparisons, even those where there was no direct evidence available.
25
However, the small number of relevant studies for some comparisons, statistical
26
heterogeneity (within pairwise comparisons), and potential inconsistency (between pairwise
27
comparisons) within the networks means that the encouraging results for interventions such
28
as biological agents should be interpreted with caution.
AC C
21
15
ACCEPTED MANUSCRIPT
1 In order to answer the question of which is the optimum treatment for sciatica and provide
3
generalisable findings, we were interested in the average treatment effect of each treatment
4
approach (to represent the diversity used in clinical practice). We therefore pooled clinically
5
heterogeneous studies. We used a random-effects model to pool the data, which is based
6
on the assumption that different studies assessed different, yet related, treatment effects.
7
However, included studies also varied in study design and risk of bias (methodological
8
diversity). There was considerable (I2 ≥75%)46 statistically significant between-study
9
heterogeneity present for a number of comparisons within the pairwise meta-analyses,
SC
RI PT
2
especially in the analyses of pain intensity, and it was not possible to ascertain how much
11
was due to clinical or methodological diversity. This needs to be taken into consideration in
12
future work.
M AN U
10
13
The network meta-analyses relied on the key assumption that the relative treatment effect of
15
one treatment versus another is the same across the entire set of studies.18,44 The use of
16
random-effects models meant that it was assumed that the common distribution of effects
17
was the same across all sets of studies. A further assumption made in the analyses was that
18
the relative efficacy of different treatments is the same at different stages in the care
19
pathway. Pragmatically, sciatica is often treated with a stepped care approach starting with
20
conservative treatments, such as non-opioid medication, progressing if necessary to more
21
invasive treatments such as epidural injections or surgery. This means that the population of
22
patients treated with conservative treatments was likely to differ from those treated with
23
invasive treatments, resulting in confounding and inconsistency within the network. Although
24
descriptive characteristics were generally poorly reported by included studies, there was a
25
trend for studies evaluating invasive treatments to report a history of previous treatments
26
and include patients with a diagnosis confirmed by imaging, and for studies of conservative
27
treatments to limit inclusion to patients with acute sciatica. Due to the breadth of the review
28
and the novel and speculative use of network meta-analysis methods, we have not yet
AC C
EP
TE D
14
16
ACCEPTED MANUSCRIPT
1
incorporated stepped care approaches in the network meta-analyses. The optimum
2
sequence of treatment modalities and what sequence is best for which patients is therefore
3
not yet known and awaits further analysis.
RI PT
4 The network meta-analyses were based on a single time point, outcome data closest to six
6
months, which may be considered as a limitation of the analyses. The HTA monograph16
7
included an assessment of each treatment strategy at short (≤ 6 weeks), medium (> 6weeks
8
to ≤ 6 months) and long (> 6 months) term follow-up, but this evaluation was based on
9
multiple pairwise analyses, with each analysis needing to be interpreted independently.
10
Further research is needed to incorporate multiple time points within the network meta-
11
analyses in order to incorporate data at different follow-up periods.
M AN U
SC
5
12
For the pain intensity outcome, where the SDs were missing (and could not be estimated
14
from the published data) these were imputed using the weighted mean SD39,175 for each
15
treatment strategy (11 studies). This is based on the assumption that the variance is similar
16
between studies and the data are not skewed.176 We also used medians to represent the
17
mean for two studies. We considered that it was better to use these methods in order to
18
incorporate more of the evidence base, as ignoring the findings of these studies may induce
19
bias in the summary effect estimate.175 Furukawa, et al.39 have previously shown that it is
20
safe to borrow SDs from other studies.
EP
AC C
21
TE D
13
22
There were insufficient studies to explore the presence of publication or reporting bias for
23
most treatment comparisons. However, a funnel plot of studies comparing surgery and
24
chemonucleolysis showed no evidence of publication bias.16 The benefit (or effectiveness) of
25
different treatment strategies for sciatica should be considered along with potential harms.
26
Although the present paper does not report adverse effects, they are reported elsewhere.16
27 28
CONCLUSIONS
17
ACCEPTED MANUSCRIPT
The use of network meta-analyses has enabled us to provide new information on the relative
2
effectiveness of treatments for sciatica. This can help clinicians and patients in shared
3
decision making, as well as providing data for healthcare policy development. The findings
4
provide support for the effectiveness of some common therapies for sciatica such as non-
5
opioid medication, epidural injections and disc surgery. They also suggest that less
6
frequently used treatments such as manipulation and acupuncture, and experimental
7
treatments such as cytokine modulating biological agents, may be considered. The findings
8
of this review do not support the effectiveness of opioid medication, either for pain intensity
9
or global effect. Furthermore, there is no support for the effectiveness of numerous other
SC
RI PT
1
interventions such as bed rest, exercise therapy, percutaneous discectomy or traction. The
11
lack of support for education/advice should not be taken to imply that patients should not be
12
given information or advice; rather it is not an effective treatment if delivered alone.
M AN U
10
13
Further research is needed to confirm or refute these findings where we found limited
15
evidence, and to explore the impact of heterogeneity and the range of clinical questions
16
most suited for the use of network meta-analyses. There is also scope to develop more
17
sophisticated methods, such as building on the confidence profile method,173 bias-adjusted
18
results,177 or Bayesian statistics,_ENREF_160172 to incorporate information relating to
19
differences in study design or internal and external validity in the network meta-analyses, as
20
well as data on multiple follow-up periods. The issue of how best to estimate the
21
effectiveness of treatment approaches according to their order within a sequential treatment
22
pathway remains an important challenge.
24 25
EP
AC C
23
TE D
14
REFERENCES
18
ACCEPTED MANUSCRIPT
1
1.
Deyo RA, Rainville J, Kent DL. What can the history and physical examination tell us
2
about low back pain? Journal of the American Medical Association. 1992;268(6):760-
3
765. 2.
Bush K, Cowan N, Katz DE, et al. The natural history of sciatica associated with disc
RI PT
4 5
pathology; a prospective study with clinical and independent radiological follow-up.
6
Spine 1992;18:1433-1438. 3.
8
Journal of Clinical Epidemiology 2004;57:174-179. 4.
10 11
Koes BW, van Tulder MW, Peul WC. Clinical Review: Diagnosis and treatment of sciatica BMJ. 2007;334:1313-1317.
5.
SC
9
Tubach F, Beaute J, Leclerc A. Natural history and prognostic indicators of sciatica.
M AN U
7
Weber H, Holme I, Amlie E. The natural course of acute sciatica with nerve root
12
symptoms in a double-blind placebo-controlled trial evaluating the effect of piroxicam.
13
Spine 1993;18:1433-1438. 6.
15 7.
17 18
8.
9.
Vroomen PC, de Krom MC, Slofstra PD, Knottnerus JA. Conservative treatment of sciatica: a systematic review. J Spinal Disord. Dec 2000;13(6):463-469.
10.
23 24
Konstantinou K, Dunn KM. Sciatica: Review of epidemiological studies and prevalence estimates. Spine 2008;33:2464-2472.
21 22
Samanta A, Beardsley J. Evidence based case report: Sciatica: which intervention? . BMJ 1999;319:302-303.
19 20
TE D
canal. . N Engl J Med 1934;211:210-215.
EP
16
Mixer WJ, Barr JS. Rupture of the intervertebral disc with involvement of the spinal
AC C
14
Hagen KB, Hilde G, Jamtvedt G, Winnem M. Best rest for acute low back pain and sciatica. The Cochrane Database of Systematic Reviews 2004.
11.
Luijsterburg PAJ, Verhagen AP, Ostelo RWJG, van Os TAG, Peul WC, Koes BW.
25
Effectiveness of conservative treatments for the lumbosacral radicular syndrome: a
26
systematic review. European Spine Journal. Jul 2007;16(7):881-899.
27 28
12.
Clarke JA, van Tulder MW, Blomberg SEI, et al. Traction for low-back pain with or without sciatica. Cochrane Database of Systematic Reviews. 2007(2).
19
ACCEPTED MANUSCRIPT
1
13.
Boswell MV, Hansen HC, Trescot AM, Hirsch JA. Epidural steroids in the
2
management of chronic spinal pain and radiculopathy. Pain Physician. Jul
3
2003;6(3):319-334.
5
spinal pain: a systematic review. Pain Physician. Jan 2007;10(1):185-212. 15.
7 8
Database of Systematic Reviews. 2007(2). 16.
9
Lewis R, Williams N, Matar HE, et al. The effectiveness and cost effectiveness of
management strategies for sciatica: systematic review and economic model. Health
10 11
Gibson JNA, Waddell G. Surgical interventions for lumbar disc prolapse. Cochrane
SC
6
Abdi S, Datta S, Trescot AM, et al. Epidural steroids in the management of chronic
RI PT
14.
Technology Assessment. 2011;15(39). 17.
M AN U
4
Psaty BM, Lumley T, Furberg CD, et al. Health Outcomes Associated With Various
12
Antihypertensive Therapies Used as First-Line Agents: A Network Meta-analysis
13
Jama. 2003;289(19):2534-2544 18.
15 16
Lu G, Ades AE. Combination of direct and indirect evidence in mixed treatment comparisons. Statistics in Medicine. 2004;23(15449338):3105-3124.
19.
TE D
14
Cooper NJ, Sutton AJ, Lu G, Khunti K. Mixed Comparison of Stroke Prevention
17
Treatments in Individuals With Nonrheumatic Atrial Fibrillation Archives of Internal
18
Medicine. 2006;166:1269-1275. 20.
Anothaisintawee T, Attia J, Nickel JC, et al. Management of Chronic Prostatitis/
EP
19
Chronic Pelvic Pain Syndrome: A Systematic Review and Network Meta-analysis
21
JAMA. 2011;305 (1):78-86.
22
21.
23 24
27
Higgins JPT, Whitehead A. Borrowing strength from external trials in a meta-analysis. Statistics in Medicine 1996;15:2733-2749.
22.
25 26
AC C
20
Sutton AJ, Abrams KR. Bayesian methods in meta-analyis and evidence synthesis. Statistical Methods in Medical Research. 2001;10:277-303.
23.
Bland MJ, Altman DG. Statistics notes: Bayesians and frequentists. BMJ. October 24 1998;317(166):1151–1160.
20
ACCEPTED MANUSCRIPT
1
24.
2 3
Ashby D, Smith AFM. Evidence-based medicine as Bayesian decision-making. Statistics in Medicine. 2000;19:3291-3305.
25.
Salanti G, Ades AE, Ioannidis JPA. Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial.
5
Journal of Clinical Epidemiology. 2011;64:163-171.
6
26.
RI PT
4
Tajiri K, Takahashi K, Ikeda K, Tomita K. Common peroneal nerve block for sciatica. Clinical Orthopaedics and Related Research. 1998(347):203-207.
8
http://www.mrw.interscience.wiley.com/cochrane/clcentral/articles/952/CN-
9
00148952/frame.html 27.
11 12
Hedges LV, Tipton E, Johnson ML. Robust variance estimation in meta-regression
M AN U
10
SC
7
with dependent effect size estimates. Research Synthesis Methods. 2010;1:39-65. 28.
13
Gibson T, Dilke TF, Grahame R. Chymoral in the treatment of lumbar disc prolapse. Rheumatology & Rehabilitation. Aug 1975;14(3):186-190.
29.
Bland M. Meta-analysis: publication bias. 2006.
15
30.
El-Zahaar MS. The Egyptian experience in the use of colchicine in lumbar disc
TE D
14
16
disease. Journal of Neurological and Orthopaedic Medicine and Surgery.
17
1995;16(3):147-152.
19 20
165. 32.
21
radiculopathy. Spine. Jun 1 2007;32(13):1414-1422.
33.
24
Lavyne MH, Bilsky MH. Epidural steroids, postoperative morbidity, and recovery in patients undergoing microsurgical lumbar discectomy. Journal of Neurosurgery. Jul
25 26
Thomas KC, Fisher CG, Boyd M, Bishop P, Wing P, Dvorak MF. Outcome evaluation of surgical and nonsurgical management of lumbar disc protrusion causing
22 23
Gelman A. Bayesian Statistics Then and Now. Statistical Science. 2010;25(2):162-
EP
31.
AC C
18
1992;77(1):90-95. 34.
Ejeskar A, Nachemson A, Herberts P, et al. Surgery versus chemonucleolysis for
27
herniated lumbar discs. A prospective study with random assignment. Clin Orthop.
28
Apr 1983(174):236-242.
21
ACCEPTED MANUSCRIPT
1
35.
van Tulder MW, Furlan A, Bombardier C, Bouter LM, Editorial Board of the Cochrane Collaboration Back Review Group. Updated method guidelines for systematic
3
reviews in the Cochrane collaboration back review group. Spine. 2003;28(12):1290-
4
1299.
5
36.
RI PT
2
Jackson N, Waters E, For the Guidelines for Systematic Reviews in Health
Promotion and Public health Taskforce. Criteria for the systematic review of health
7
promotion and public health interventions. Health Promotion International.
8
2005;20(4):367-374. 37.
10 11
Effective Public Health Practice Project. Quality Assessment tool for Quantitative Studies http://www.ephpp.ca/tools.html. 2007 (Accessed 31 July, 2007).
38.
M AN U
9
SC
6
Higgins JPT, Deeks JJ. Chapter 7: Selecting studies and collecting data. In: Higgins
12
JPT, Green S, eds. Cochrane handbook for systematic reviews of interventions.
13
Chichester (UK): John Wiley & Sons Ltd; 2008.
14
39.
Furukawa TA, Barbui C, Cipriani A, Brambilla P, Watanabe N. Imputing missing standard deviations in meta-analyses can provide accurate results. J Clin Epidemiol.
16
Jan 2006;59(1):7-10. 40.
18
manual. Cambridge: MRC Biostatistics Unit; 2003. 41.
20 21
1986;7:177-188. 42.
22
Song F, Altman DG, Glenny A-M, Deeks JJ. Validity of indirect comparison for estimating efficacy of competing interventions: empirical evidence from published
23 24
DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled clinical trials
EP
19
Spiegelhalter DJ, Thomas A, Best NG, Lunn DJ, eds. WinBUGS Version v1.4. User
AC C
17
TE D
15
meta-analyses. BMJ. 2003;326:472-476.
43.
Bucher HC, Guyatt GH, Griffith LE, Walter SD. The Results of Direct and Indirect
25
Treatment Comparisons in Meta-Analysis of Randomized Controlled Trials. J Clin
26
Epidemiol. 1997;50(6):683-691.
22
ACCEPTED MANUSCRIPT
1
44.
Caldwell DM, Ades AE, Higgins JPT. Simultaneous comparison of multiple
2
treatments: combining direct and indirect evidence. BMJ. 2005;331(16223826):897-
3
900. 45.
Stettler C, Allemann S, Wandel S, et al. Drug eluting and bare metal stents in people
RI PT
4 5
with and without diabetes: collaborative network meta-analysis BMJ.
6
2008;337:a1331.
7
46.
Deeks JJ, Higgins JPT, Altman DG. Chapter 9: Analysing data and undertaking
meta-analyses. In: Higgins JPT, Green S, eds. Cochrane Handbook for Systematic
9
Reviews of Interventions. Chichester (UK): John Wiley & Sons; 208. 47.
11 12
Higgins JPT, Thompson SG, Deeks JJ, G. AD. Measuring inconsistency in meta-
M AN U
10
SC
8
analyses. BMJ. 2003;327(7414):557-560 48.
Alexander AH, Burkus JK, Mitchell JB, Ayers WV. Chymopapain chemonucleolysis
13
versus surgical discectomy in a military population. Clin Orthop Relat Res.
14
1989(244):158-165. 49.
Aminmansour B, Khalili HA, Ahmadi J, Nourian M. Effect of high-dose intravenous
TE D
15 16
dexamethasone on postlumbar discectomy pain. Spine. Oct 1 2006;31(21):2415-
17
2417.
18
50.
Atlas SJ, Deyo RA, Keller RB, et al. The Maine Lumbar Spine Study .2. 1-year outcomes of surgical and nonsurgical management of sciatica. Spine. Aug
20
1996;21(15):1777-1786. 51.
22
Efficacy of epidural perineural injections with autologous conditioned serum for
23
lumbar radicular compression: an investigator-initiated, prospective, double-blind,
24 25
Becker C, Heidersdorf S, Drewlo S, de Rodriguez SZ, Kramer J, Willburger RE.
AC C
21
EP
19
reference-controlled study. Spine. Aug 1 2007;32(17):1803-1808.
52.
Bernsmann K, Kramer J, Ziozios I, Wehmeier J, Wiese M. Lumbar micro disc surgery
26
with and without autologous fat graft - A prospective randomized trial evaluated with
27
reference to clinical and social factors. Arch. Orthop. Trauma Surg. Sep
28
2001;121(8):476-480.
23
ACCEPTED MANUSCRIPT
1
53.
Bonafe A, Tremoulet M, Sabatier J, et al. Foraminal and Lateroforaminal Herniated
2
Disks - Midterm Results of Percutaneous Treatment with Chemonucleolysis or
3
Nucleotomy. Neurochirurgie. 1993;39(2):110-115. 54.
Bontoux D, Alcalay M, Debiais F, et al. Treatment of Lumbar-Disk Herniation with
RI PT
4 5
Intradiscal Injection of Chymopapain or Triamcinolone Hexacetonid - a Comparative-
6
Study of 80 Cases. Rev Rhum Engl Ed. Apr 1990;57(4):327-331.
7
55.
Bourgeois P, Benoist M, Palazzo E, et al. Multicenter, randomized and double-blind study of triamcinolone hexacetonide versus chymopapain in the treatment of lumbar
9
sciatica. Preliminary results at six months. Rev Rhum Mal Osteo Articulaires.
10 56.
12 13 14 15
Brown MD, Tompkins JS. Pain response post-chemonucleolysis or disc excision. Spine. Mar 1989;14(3):321-326.
57.
M AN U
1988(10):767-769.
Buchner M, Zeifang F, Brocai DR, Schiltenwolf M. Epidural corticosteroid injection in the conservative management of sciatica. Clin Orthop. Jun 2000(375):149-156.
58.
Buric J. Ozone chemyonucleolysis vs microdiscectomy. Prospective controlled study
TE D
11
SC
8
16
with 18 months follow-up. Rivista Italiana di Ossigeno-Ozonoterapia. Apr
17
2005;4(1):49-54.
18
59.
Burton AK, Tillotson KM, Cleary J. Single-blind randomised controlled trial of chemonucleolysis and manipulation in the treatment of symptomatic lumbar disc
20
herniation. European Spine Journal. Jun 2000;9(3):202-207. 60.
22
plus procaine for the management of intractable sciatica. Spine. May 1991;16(5):572-
23 24
Bush K, Hillier S. A controlled study of caudal epidural injections of triamcinolone
AC C
21
EP
19
575.
61.
Carette S, Leclaire R, Marcoux S, et al. Epidural corticosteroid injections for sciatica
25
due to herniated nucleus pulposus. New England Journal of Medicine. Jun
26
1997;336(23):1634-1640.
27 28
62.
Cengiz SL, Baysefer A. Efficacy of Adcon-L gel or Healon-GV in epidural fibrosis after lumbar microdiscectomy. Neurosciences. Apr 2007;12(2):109-113.
24
ACCEPTED MANUSCRIPT
1
63.
Chatterjee S, Foy PM, Findlay GF. Report of a controlled clinical trial comparing
2
automated percutaneous lumbar discectomy and microdiscectomy in the treatment of
3
contained lumbar disc herniation. Spine. Mar 15 1995;20(6):734-738. 64.
Chen M-r, Wang P, Cheng G, Guo X, Wei G-w, Cheng X-h. The warming
RI PT
4 5
acupuncture for treatment of sciatica in 30 cases. J Tradit Chin Med. 2009;29(1):50-
6
53.
7
65.
Cohen S, Bogduk N, Dragovich A, et al. Randomized, double-blind, placebo-
controlled, dose-response, and preclinical safety study of transforaminal epidural
9
etanercept for the treatment of sciatica. Anesthesiology. 2009;110(5):1116-1126. 66.
11 12
Coomes EN. Comparison between Epidural Anaesthesia and Bed Rest in Sciatica.
M AN U
10
SC
8
British Medical Journal. 1961;1(5218):20-24. 67.
Crawshaw C, Frazer AM, Merriam WF, Mulholland RC, Webb JK. A comparison of
13
surgery and chemonucleolysis in the treatment of sciatica. A prospective randomized
14
trial. Spine. Mar 1984;9(2):195-198. 68.
16 17
Dabezies EJ, Brunet M. Chemonucleolysis vs laminectomy. Orthopedics. Jan-Feb 1978;1(1):26-29.
69.
TE D
15
Dabezies EJ, Langford K, Morris J, Shields CB, Wilkinson HA. Safety and efficacy of chymopapain (Discase) in the treatment of sciatica due to a herniated nucleus
19
pulposus. Results of a randomized, double-blind study. Spine. May 1988;13(5):561-
20
565. 70.
22 23
discectomy. J Spinal Disord Tech. Aug 2002;15(4):273-276.
71.
24 25
Debi R, Halperin N, Mirovsky Y. Local application of steroids following lumbar
AC C
21
EP
18
Dei-Anang K, Weigand H, Mader U. Is the Percutaneous Discectomy an Alternative to Chemonucleolysis. Radiologe. 1990;30(2):70-74.
72.
Dilke TF, Burry HC, Grahame R. Extradural corticosteroid injection in management of
26
lumbar nerve root compression. British Medical Journal. Jun 16 1973;2(5867):635-
27
637.
25
ACCEPTED MANUSCRIPT
1
73.
Dincer U, Kiralp MZ, Cakar E, Yasar E, Dursan H. Caudal epidural injection versus
2
non-steroidal anti-inflammatory drugs in the treatment of low back pain accompanied
3
with radicular pain. Joint Bone Spine. Oct 2007;74(5):467-471. 74.
Dreiser RL, Le Parc JM, Velicitat P, Lleu PL. Oral meloxicam is effective in acute
RI PT
4 5
sciatica: two randomised, double-blind trials versus placebo or diclofenac.
6
Inflammation Research. Mar 2001;50 Suppl 1:S17-23. 75.
8 9
Dubourg G, Rozenberg S, Fautrel B, et al. A pilot study on the recovery from paresis after lumbar disc herniation. Spine. Jul 1 2002;27(13):1426-1431; discussion 1431.
76.
SC
7
Duplan B, Cabanel G, Piton JL, Grauer JL, Phelip X. Acupuncture and sciatica in the acute phase. Double-blind study of 30 cases. Semaine des Hopitaux. Dec 8
11
1983;59(45):3109-3114.
12
77.
M AN U
10
Feldman J, Menkes CJ, Pallardy G, et al. Double-blind study of the treatment of disc
13
lumbosciatica by chemonucleolysis. Rev Rhum Mal Osteoartic. Mar 1986;53(3):147-
14
152. 78.
Finckh A, Zufferey P, Schurch M-A, Balague F, Waldburger M, So AKL. Short-term
TE D
15 16
efficacy of intravenous pulse glucocorticoids in acute discogenic sciatica. A
17
randomized controlled trial. Spine. Feb 15 2006;31(4):377-381.
18
79.
Gallucci M, Limbucci N, Zugaro L, et al. Sciatica: treatment with intradiscal and intraforaminal injections of steroid and oxygen-ozone versus steroid only. Radiology.
20
Mar 2007;242(3):907-913. 80.
22 23
severe sciatica: a pilot study. Ann Rheum Dis. Sep 2004;63(9):1120-1123.
81.
24
Gerszten PC, Moossy JJ, Flickinger JC, Welch WC. Low-dose radiotherapy for the inhibition of peridural fibrosis after reexploratory nerve root decompression for
25 26
Genevay S, Stingelin S, Gabay C. Efficacy of etanercept in the treatment of acute,
AC C
21
EP
19
postlaminectomy syndrome. Journal of Neurosurgery. Oct 2003;99(3 Suppl):271-277. 82.
Geurts JWM, van Wijk R, Wynne HJ, et al. Radiofrequency lesioning of dorsal root
27
ganglia for chronic lumbosacral radicular pain: a randomised, double-blind, controlled
28
trial. Lancet. Jan 2003;361(9351):21-26.
26
ACCEPTED MANUSCRIPT
1
83.
Ghoname EA, White PF, Ahmed HE, Hamza MA, Craig WF, Noe CE. Percutaneous
2
electrical nerve stimulation: an alternative to TENS in the management of sciatica.
3
Pain. Nov 1999;83(2):193-199. 84.
Glasser RS, Knego RS, Delashaw JB, Fessler RG. The perioperative use of
RI PT
4 5
corticosteroids and bupivacaine in the management of lumbar disc disease. Journal
6
of Neurosurgery. Mar 1993;78(3):383-387. 85.
8
1992;17(4):388-394. 86.
10 11
Goldie I. A clinical trial with indomethacin (indomee(R)) in low back pain and sciatica. Acta Orthop Scand Suppl. 1968;39(1):117-128.
87.
SC
9
Gogan WJ, Fraser RD. Chymopapain - a 10-Year, Double-Blind-Study. Spine. Apr
M AN U
7
Graham CE. Chemonucleolysis: a double blind study comparing chemonucleolysis
12
with intra discal hydrocortisone: in the treatment of backache and sciatica. Clin
13
Orthop. Jun 1976(117):179-192. 88.
15 16
Zeitschrift fur Rheumatologie. Nov-Dec 1975;34(11-12):444-447. 89.
17 18
Grevsten S, Johansson H. Phenylbutazone in treatment of acute lumbago-sciatica.
TE D
14
Hansson E, Hansson T. The cost-utility of lumbar disc herniation surgery. European Spine Journal. Mar 2007;16(3):329-337.
90.
Hedeboe J, Buhl M, Ramsing P. Effects of using dexamethasone and placebo in the treatment of prolapsed lumbar disc. Acta Neurologica Scandinavica. Jan
20
1982;65(1):6-10. 91.
22
sciatica by a single extradural corticosteroid injection. Br J Clin Pract.
23 24
Helliwell M, Robertson JC, Ellis RM. Outpatient treatment of low pack pain and
AC C
21
EP
19
1985;39(6):228-231.
92.
Herrmann WA, Geertsen MS. Efficacy and safety of lornoxicam compared with
25
placebo and diclofenac in acute sciatica/lumbo-sciatica: an analysis from a
26
randomised, double-blind, multicentre, parallel-group study. Int J Clin Pract.
27
2009;63(11):1613-1621.
27
ACCEPTED MANUSCRIPT
1
93.
Hofstee DJ, Gijtenbeek JMM, Hoogland PH, et al. Westeinde sciatica trial:
2
randomized controlled study of bed rest and physiotherapy for acute sciatica. Journal
3
of Neurosurgery. Jan 2002;96(1 Suppl):45-49. 94.
5 6
Holve RL, Barkan H. Oral steroids in initial treatment of acute sciatica. J Am Board
RI PT
4
Fam Med. 2008;21(5):469-474. 95.
Hoogmartens M, Van Loon L, Demedts D, Desmet L. Comparative study of the
results obtained with discectomy and chemonucleolysis. Acta Orthop Belg. Mar-Apr
8
1976;42(2):134-138.
9
96.
SC
7
Javid MJ, Nordby EJ, Ford LT, et al. Safety and efficacy of chymopapain (Chymodiactin) in herniated nucleus pulposus with sciatica. Results of a randomized,
11
double-blind study. Journal of the American Medical Association. May 13
12
1983;249(18):2489-2494.
13
97.
14
Javid MJ. Chemonucleolysis Versus Laminectomy - a Cohort Comparison of Effectiveness and Charges. Spine. Sep 1995;20(18):2016-2022.
98.
Jensen TT, Asmussen K, Berg-Hansen EM, et al. First-time operation for lumbar disc
TE D
15
M AN U
10
16
herniation with or without free fat transplantation. Prospective triple-blind randomized
17
study with reference to clinical factors and enhanced computed tomographic scan 1
18
year after operation. Spine. May 1 1996;21(9):1072-1076. 99.
Jirarattanaphochai K, Jung S, Thienthong S, Krisanaprakornkit W, Sumananont C.
EP
19
Peridural methylprednisolone and wound infiltration with bupivacaine for
21
postoperative pain control after posterior lumbar spine surgery - A randomized
22 23
double-blinded placebo-controlled trial. Spine. 2007;32(6):609-616.
100.
24 25
AC C
20
Karppinen J, Malmivaara A, Kurunlahti M, et al. Periradicular infiltration for sciatica A randomized controlled trial. Spine. May 2001;26(9):1059-1067.
101.
Karppinen J, Korhonen T, Malmivaara A, et al. Tumor necrosis factor-alpha
26
monoclonal antibody, infliximab, used to manage severe sciatica. Spine. Apr 15
27
2003;28(8):750-753; discussion 753-754.
28
ACCEPTED MANUSCRIPT
102.
2 3
chronic lumbar radicular pain. The Journal of Pain. 2005;6(12):829-836. 103.
4 5
Khoromi S, Patsalides A, Parada S, Salehi V, Meegan JM, Max MB. Topiramate in
Khoromi S, Cui L, Nackers L, Max MB. Morphine, nortriptyline and their combination vs. placebo in patients with chronic lumbar root pain. Pain. Jul 2007;130(1-2):66-75.
104.
RI PT
1
Kim KD, Wang JC, Robertson DP, et al. Reduction of radiculopathy and pain with
6
Oxiplex/SP gel after laminectomy, laminotomy, and discectomy: a pilot clinical study.
7
Spine. May 15 2003;28(10):1080-1087; discussion 1087-1088.
9
endoscopic diskectomy in 295 patients: comparison with results of microscopic
10
diskectomy. Surg Neurol. Dec 2007;68(6):623-631. 106.
12 13
Klenerman L, Greenwood R, Davenport HT, White DC, Peskett S. Lumbar epidural injections in the treatment of sciatica. Br. J. Rheumatol. Feb 1984;23(1):35-38.
107.
14 15
M AN U
11
Kim M-J, Lee S-H, Jung E-S, et al. Targeted percutaneous transforaminal
SC
105.
Koranda I, Sefrna F. Effectiveness of conservative and surgical treatment of lumboischiadic syndrome. Cas Lek Cesk. Aug 2 1995;134(15):474-477.
108.
Korhonen T, Karppinen J, Paimela L, et al. The treatment of disc herniation-induced
TE D
8
16
sciatica with infliximab - Results of a randomized, controlled, 3-month follow-up
17
study. Spine. Dec 2005;30(24):2724-2728.
18
109.
Kwasucki J, Olbrych K, ska B, Stepie A, Dec S, Maksymiuk G. Assessment of dexamethasone effectiveness in the treatment of ischalgia. Neurol Neurochir Pol.
20
1993;27(4):515-522. 110.
22
action of fluvoxamine compared with efficacy of imipramine and tramadol for
23 24
Kwasucki J, Stepien A, Maksymiuk G, Olbrych-Karpinska B. Evaluation of analgesic
AC C
21
EP
19
treatment of sciatica--open trial. Wiad Lek. 2002;55(1-2):42-50.
111.
Lagarrigue J, Lazorthes Y, Verdie JC, Richaud J. Comparative Results of Surgery
25
and Nucleolysis in 1085 Cases of Herniated Lumbar Disc. Neurochirurgie.
26
1991;37(2):96-105.
29
ACCEPTED MANUSCRIPT
1
112.
Laiq N, Khan MN, Iqbal MJ, Khan S. Comparison of Epidural Steroid Injections with
2
conservative management in patients with lumbar radiculopathy. J Coll Physicians
3
Surg Pak. 2009;19(9):539-543. 113.
Larsson U, Choler U, Lidstrom A, et al. Auto Traction for Treatment of Lumbago
RI PT
4 5
Sciatica a Multicenter Controlled Investigation. Acta Orthop Scand Suppl.
6
1980;51(5):791-798.
7
114.
Lavignolle B, Vital JM, Baulny D, Grenier F, Castagnera L. Comparative study of
surgery and chemonucleolysis in the treatment of sciatica caused by a herniated
9
disk. Acta Orthop Belg. 1987;53(2):244-249. 115.
Lee SH, Lee SJ, Park KH, et al. Comparison of percutaneous manual and
M AN U
10
SC
8
11
endoscopic laser diskectomy with chemonucleolysis and automated nucleotomy.
12
Orthopade. 1996;25(1):49-55.
13
116.
Lee SH, Chung SE, Ahn Y, Kim TH, Park JY, Shin SW. Comparative radiologic evaluation of percutaneous endoscopic lumbar discectomy and open
15
microdiscectomy: A matched cohort analysis. Mount Sinai Journal of Medicine. Sep
16
2006;73(5):795-801. 117.
18 19
Lidstrom A, Zachrisson M. Physical therapy on low back pain and sciatica. An attempt at evaluation. Scand J Rehabil Med. 1970;2(1):37-42.
118.
Ljunggren AE, Walker L, Weber H, Amundsen T. Manual traction versus isometric
EP
17
TE D
14
exercises in patients with herniated intervertebral lumbar discs. Physiotherapy
21
Theory and Practice. Dec 1992;8(4):207-213.
22
119.
23
Luijsterburg PAJ, Verhagen AP, Ostelo RWJG, et al. Physical therapy plus general practitioners' care versus general practitioners' care alone for sciatica: a randomised
24
clinical trial with a 12-month follow-up. European Spine Journal. Apr 2008;17(4):509-
25 26
AC C
20
517. 120.
Lundin A, Magnuson A, Axelsson K, Kogler H, Samuelsson L. The effect of
27
perioperative corticosteroids on the outcome of microscopic lumbar disc surgery.
28
European Spine Journal. Dec 2003;12(6):625-630.
30
ACCEPTED MANUSCRIPT
1
121.
Mackay MA, Fischgrund JS, Herkowitz HN, Kurz LT, Hecht B, Schwartz M. The
2
Effect of Interposition Membrane on the Outcome of Lumbar Laminectomy and
3
Diskectomy. Spine. Aug 1995;20(16):1793-1796. 122.
Mathews JA, Mills SB, Jenkins VM, et al. Back pain and sciatica: controlled trials of
RI PT
4 5
manipulation, traction, sclerosant and epidural injections. Br J Rheumatol.
6
1987;26(6):416-423.
7
123.
Mayer HM, Brock M. Percutaneous Endoscopic Discectomy - Surgical Technique and Preliminary-Results Compared to Microsurgical Discectomy. Journal of
9
Neurosurgery. Feb 1993;78(2):216-225. 124.
Moret NC, van der Stap M, Hagmeijer R, Molenaar A, Koes BW. Design and
M AN U
10
SC
8
11
feasibility of a randomized clinical trial to evaluate the effect of vertical traction in
12
patients with a lumbar radicular syndrome. Manual Therapy. 1998;3(4):203-211.
13
125.
Muralikuttan KP, Hamilton A, Kernohan WG, Mollan RA, Adair IV. A prospective randomized trial of chemonucleolysis and conventional disc surgery in single level
15
lumbar disc herniation. Spine. Apr 1992;17(4):381-387.
16
126.
TE D
14
Murata Y, Kato Y, Miyamoto K, Takahashi K. Clinical study of low back pain and
17
radicular pain pathways by using l2 spinal nerve root infiltration: a randomized,
18
controlled, clinical trial. Spine. 2009;34(19):2008-2013. 127.
North RB, Kidd DH, Farrokhi F, Piantadosi SA. Spinal cord stimulation versus
EP
19
repeated lumbosacral spine surgery for chronic pain: a randomized, controlled trial.
21
Neurosurgery. 2005;56(1):98-106; discussion 106-107.
22
128.
23 24
AC C
20
Norton WL. Chemonucleolysis versus surgical discectomy. Comparison of costs and results in workers' compensation claimants. Spine. Jun 1986;11(5):440-443.
129.
Osterman H, Seitsalo S, Karppinen J, Malmivaara A. Effectiveness of
25
microdiscectomy for lumbar disc herniation: a randomized controlled trial with 2 years
26
of follow-up. Spine. Oct 1 2006;31(21):2409-2414.
31
ACCEPTED MANUSCRIPT
1
130.
Ozturk B, Gunduz OH, Ozoran K, Bostanoglu S. Effect of continuous lumbar traction
2
on the size of herniated disc material in lumbar disc herniation. Rheumatology
3
International. May 2006;26(7):622-626. 131.
Peul WC, Van Houwelingen HC, Van Den Hout WB, et al. Surgery versus prolonged
RI PT
4 5
conservative treatment for sciatica. New England Journal of Medicine. 31
6
2007;356(22):2245-2256.
7
132.
Popiolek A, Domanik A, Mazurkiewicz G. Epidural injections of steroids in the
treatment of patients with chronic sciatica in discopathy. Neurol Neurochir Pol. Sep-
9
Oct 1991;25(5):640-646. 133.
Porsman O, Friis H. Prolapsed lumbar disc treated with intramuscularly administered
M AN U
10
SC
8
11
dexamethasonephosphate. A prospectively planned, double-blind, controlled clinical
12
trial in 52 patients. Scandinavian Journal of Rheumatology. 1979;8(3):142-144.
13
134.
Postacchini F, Lami R, Massobrio M. Chemonucleolysis versus surgery in lumbar disc herniations: correlation of the results to preoperative clinical pattern and size of
15
the herniation. Spine. Mar 1987;12(2):87-96.
16
135.
TE D
14
Price C, Arden N, Coglan L, Rogers P. Cost-effectiveness and safety of epidural
17
steroids in the management of sciatica. Health Technology Assessment. Aug
18
2005;9(33):1-58. 136.
Rasmussen S, Krum-Moller DS, Lauridsen LR, et al. Epidural steroid following
EP
19
discectomy for herniated lumbar disc reduces neurological impairment and enhances
21
recovery: a randomized study with two-year follow-up. Spine. 2008;33(19):2028-
22 23
AC C
20
2033.
137.
24
Rattanatharn R, Sanjaroensuttikul N, Anadirekkul P, Chaivisate R, Wannasetta W. Effectiveness of lumbar traction with routine conservative treatment in acute
25
herniated disc syndrome. Journal of the Medical Association of Thailand. Sep
26
2004;87 Suppl 2:S272-277.
27 28
138.
Reust P, Chantraine A, Vischer TL. Treatment of lumbar sciatica with or without neurological deficit using mechanical traction. A double-blind study. Schweizerische
32
ACCEPTED MANUSCRIPT
1
Medizinische Wochenschrift. Feb 27 1988;Journal Suisse de Medecine. 118(8):271-
2
274.
3
139.
Revel M, Payan C, Vallee C, et al. Automated Percutaneous Lumbar Discectomy Versus Chemonucleolysis in the Treatment of Sciatica - a Randomized Multicenter
5
Trial. Spine. Jan 1993;18(1):1-7.
6
140.
RI PT
4
Richter HP, Kast E, Tomczak R, Besenfelder W, Gaus W. Results of applying
ADCON-L gel after lumbar discectomy: the German ADCON-L study. Journal of
8
Neurosurgery. Oct 2001;95(2 Suppl):179-189.
9
141.
SC
7
Ridley MG, Kingsley GH, Gibson T, Grahame R. Outpatient lumbar epidural corticosteroid injection in the management of sciatica. Br. J. Rheumatol. Aug
11
1988;27(4):295-299.
12
142.
M AN U
10
Ronnberg K, Lind B, Zoega B, et al. Peridural scar and its relation to clinical
13
outcome: a randomised study on surgically treated lumbar disc herniation patients.
14
European Spine Journal. 2008;17(12):1714-1720. 143.
Santilli V, Beghi E, Finucci S. Chiropractic manipulation in the treatment of acute
TE D
15 16
back pain and sciatica with disc protrusion: a randomized double-blind clinical trial of
17
active and simulated spinal manipulations. Spine J. Mar-Apr 2006;6(2):131-137.
18
144.
Schwetschenau PR, Ramirez A, Johnston J. Double blind evaluation of intradiscal chymopapain for herniated lumbar discs; early results. J Neurosurg. 1976;45(6):622-
20
627. 145.
22
analysis of extended conservative therapy versus surgical intervention in the
23 24
Shvartzman L, Weingarten E, Sherry H, Levin S, Persaud A. Cost-effectiveness
AC C
21
EP
19
management of herniated lumbar intervertebral disc. Spine. Feb 1992;17(2):176-182.
146.
Snoek W, Weber H, Jorgensen B. Double-Blind Evaluation of Extradural Methyl
25
Prednisolone for Herniated Lumbar Discs. Acta Orthop Scand Suppl.
26
1977;48(6):635-641.
33
ACCEPTED MANUSCRIPT
1
147.
Steffen R. Laser discotomy/Chemonucleolysis ini the therapy of displaced lumbal
2
intervertebral disc - prospective randomised comparison. Hefte zur der Unfallchirurg.
3
1999:215-221. 148.
Stula D. Chemonucleolysis with Chymopapain in Intervertebral-Disk Hernia - a
RI PT
4 5
Randomized Comparative-Study of Patients Following Surgery. Neurochirurgia
6
(Stuttg). Sep 1990;33(5):169-172.
7
149.
Styczynski T, Zarski S, Drozdowska Z, et al. Clinical Evaluation of the Effectiveness of Traction Treatment of Patients with Herniation of Lumbar Intervertebral Discs.
9
Reumatologia (Warsaw). 1991;29(3-4):275-281. 150.
Tassi GP. Comparison of results of 500 microdiscectomies and 500 percutaneous
M AN U
10
SC
8
11
laser disc decompression procedures for lumbar disc herniation. Photomed Laser
12
Surg. Dec 2006;24(6):694-697.
13
151.
Tregonning GD, Transfeldt EE, McCulloch JA, Macnab I, Nachemson A. Chymopapain versus conventional surgery for lumbar disc herniation. 10-year results
15
of treatment. J Bone Joint Surg Br. May 1991;73(3):481-486.
16
152.
TE D
14
de Tribolet N, Porchet F, Lutz TW, et al. Clinical assessment of a novel antiadhesion barrier gel: prospective, randomized, multicenter, clinical trial of ADCON-L to inhibit
18
postoperative peridural fibrosis and related symptoms after lumbar discectomy. Am J
19
Orthop. 1998;27(2):111-120. 153.
21
modalities for acute pain in lumbar disc herniation measured by clinical evaluation
22
and magnetic resonance imaging. J Manipulative Physiol Ther. Mar 2008;31(3):191-
23 24
Unlu Z, Tasci S, Tarhan S, Pabuscu Y, Islak S. Comparison of 3 physical therapy
AC C
20
EP
17
198.
154.
Vad VB, Bhat AL, Lutz GE, Cammisa F. Transforaminal epidural steroid injections in
25
lumbosacral radiculopathy: a prospective randomized study. Spine. Jan 1
26
2002;27(1):11-16.
34
ACCEPTED MANUSCRIPT
1
155.
Valat JP, Giraudeau B, Rozenberg S, et al. Epidural corticosteroid injections for
2
sciatica: a randomised, double blind, controlled clinical trial. Ann Rheum Dis. Jul
3
2003;62(7):639-643. 156.
van Alphen HA, Braakman R, Bezemer PD, Broere G, Berfelo MW.
RI PT
4 5
Chemonucleolysis versus discectomy: a randomized multicenter trial. Journal of
6
Neurosurgery. Jun 1989;70(6):869-875.
7
157.
Veihelmann A, Devens C, Trouillier H, Birkenmaier C, Gerdesmeyer L, Refior HJ.
Epidural neuroplasty versus physiotherapy to relieve pain in patients with sciatica: a
9
prospective randomized blinded clinical trial. J Orthop Sci. Jul 2006;11(4):365-369. 158.
Vroomen PC, de Krom MC, Wilmink JT, Kester AD, Knottnerus JA. Lack of
M AN U
10
SC
8
11
effectiveness of bed rest for sciatica. New England Journal of Medicine. Feb 11
12
1999;340(6):418-423.
13
159.
Watters WC, Mirkovic S, Boss J. Treatment of the isolated lumbar intervertebral disc herniation: microdiscectomy versus chemonucleolysis. Spine (Phila Pa 1976).
15
1988;13(3):360-362.
16
160.
TE D
14
Watts C, Hutchison G, Stern J, Clark K. Comparison of intervertebral disc disease
17
treatment by chymopapain injection and open surgery. Journal of Neurosurgery. Apr
18
1975;42(4):397-400.
20 21
observation. Spine. Mar 1983;8(2):131-140. 162.
22
Weber H, Holme I, Amlie E. The natural course of acute sciatica with nerve root symptoms in a double-blind placebo-controlled trial evaluating the effect of piroxicam.
23 24
Weber H. Lumbar disc herniation. A controlled, prospective study with ten years of
EP
161.
AC C
19
Spine. Sep 1 1993;18(11):1433-1438.
163.
Wehling P, Reinecke J. Acupuncture together with cytokine depressing herbs in
25
comparison to injection therapy with steroids in sciatic pain. Schmerz. Jun 13
26
1997;11(3):180-184.
27 28
164.
Weinstein JN, Tosteson TD, Lurie JD, et al. Surgical vs nonoperative treatment for lumbar disk herniation: the Spine Patient Outcomes Research Trial (SPORT): a
35
ACCEPTED MANUSCRIPT
1
randomized trial. Journal of the American Medical Association. Nov 22
2
2006;296(20):2441-2450.
3
165.
Weinstein JN, Lurie JD, Tosteson TD, et al. Surgical vs nonoperative treatment for lumbar disk herniation: The Spine Patient Outcomes Research Trial (SPORT)
5
observational cohort. Journal of the American Medical Association. 22
6
2006;296(20):2451-2459.
7
166.
RI PT
4
Weinstein J, Spratt KF, Lehmann T, McNeill T, Hejna W. Lumbar disc herniation. A
comparison of the results of chemonucleolysis and open discectomy after ten years.
9
J Bone Joint Surg Am. Jan 1986;68(1):43-54. 167.
11 12
with chronic radiculopathy. The Pain Clinic. 2003;15(3):213-218. 168.
13 14
Yildirim K, Sisecioglu M, Karatay S, et al. The effectiveness of gabapentin in patients
M AN U
10
SC
8
Zeiger HE, Jr. Comparison of chemonucleolysis and microsurgical discectomy for the treatment of herniated lumbar disc. Spine. Oct 1987;12(8):796-799.
169.
Armon C, Argoff CE, Samuels J, Backonja MM. Assessment: Use of epidural steroid injections to treat radicular lumbosacral pain: Report of the Therapeutics and
16
Technology Assessment Subcommittee of the American Academy of Neurology.
17
Neurology. Mar 2007;68(10):723-729.
18
170.
TE D
15
Chou R, Atlas SJ, Stanos SP, Rosenquist RW. Nonsurgical interventional therapies for low back pain: a review of the evidence for an American Pain Society clinical
20
practice guideline. Spine 2009;34(10):1078-1093. 171.
22 23
studies. Health Technol Assess. 2003;7(27).
172.
24 25 26
Deeks JJ, Dinnes J, D'Amico R, et al. Evaluating non-randomised intervention
AC C
21
EP
19
Turner RM, Spiegelhalter DJ, Smith GCS, Thompson SG. Bias modelling in evidence synthesis. J R Stat Soc Ser A Stat Soc. 2009;172(19381328):21-47.
173.
Eddy DM, Hasselblad V, Shachter R. An introduction to Bayesian methods for Metaanalysis: The Confidence Profile Method. Medical Decision Making. 1990;10:15-23.
36
ACCEPTED MANUSCRIPT
1
174.
Egger M, Smith GD, Altman D, Schneider M. Chapter 12: Systematic reviews of
2
observational studies. In: Egger M, Smith GD, Altman D, eds. Systematic Reviews in
3
Health Care: Meta-analysis in Context. London: BMJ Books; 2001. 175.
Wiebe N, Vandermeer B, Platt RW, Klassen TP, Moher D, Barrowman NJ. A
RI PT
4 5
systematic review identifies a lack of standardization in methods for handling missing
6
variance data. J Clin Epidemiol. Apr 2006;59(4):342-353. 176.
8 9
Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions. Vol Version 5.0.2. 2 ed. Chichester (UK): John Wiley & Sons Ltd; 2009.
177.
SC
7
Thompson S, Ekelund U, Jebb S, et al. A proposed method of bias adjustment for meta-analyses of published observational studies. International Journal of
11
Epidemiology 2011;40:765–777
12
M AN U
10
Figure legends:
14
Figure 1: flow diagram showing the number of references identified, publications retrieved for
15
assessment, and studies included in the review
16
Figure 2: Network of treatment strategies for sciatica for comparative studies reporting
17
global effect
18
Figure 3: Network of treatment strategies for sciatica for comparative studies reporting pain
19
intensity
20
Figure 4: Plot of the odds ratios (ORs) of global effect for different treatment strategies
21
compared with inactive control from the network meta-analysis
22
Figure 5: Plot of the weighted mean difference for pain intensity for different treatment
23
strategies compared with inactive control from the network meta-analysis
AC C
EP
TE D
13
37
ACCEPTED MANUSCRIPT
SUPPLEMENTARY DATA Manuscript title: Comparative clinical effectiveness of management strategies for sciatica: systematic review and mixed treatment comparisons metaanalyses.
RI PT
INCLUDES:
Web appendix A (pg 3): Winbug codes.
design and outcome data for the pairwise studies. Table B1 (pg 6): Description of interventions.
M AN U
Table B2 (pg 11): Description of participants.
SC
Web appendix B (pg 6): Description of the interventions, population, study
Table B3 (pg 21): Summary of overall quality of included studies. Table B4 (pg 25): Outcome data for global effect.
Table B5 (pg 33): Outcome data for pain intensity.
Web appendix C (pg 37): Results of inactive control comparisons from network meta-analyses.
TE D
Table C1 (pg 37): Probability of being best and the ORs of global effect for different treatment strategies compared with inactive control from the network meta-analysis and pairwise meta-analyses.
Table C2 (pg 38): Probability of being best and the ORs for global effect of different treatment strategies compared with inactive control from the network meta-analysis
EP
and pairwise meta-analyses based on RCTs and Q-RCTs only. Figure C1 (pg 39): Plot of the ORs of global effects for the different treatment
AC C
strategies compared with inactive control from the network meta-analysis, based on RCTs and Q-RCTs only. Table C3 (pg 40): Probability of being best and the weighted mean difference for pain intensity for different treatment strategies compared with inactive control from the network meta-analysis and pairwise meta-analyses. Table C4 (pg 41): Probability of being best and weighted mean difference for pain intensity of different treatment strategies compared with inactive control from the network meta-analysis and pairwise meta-analyses based on RCTs and Q-RCTs only.
1
ACCEPTED MANUSCRIPT
Figure C2 (pg 42): Plot of the weighted mean difference for pain intensity for the different treatment strategies compared with inactive control from the MTC analysis, based on RCTs and Q-RCTs only.
Web appendix D (pg 43): Results of network meta-analyses restricted to RCTs
RI PT
and Q-RCTs.
Table D1 (pg 43): Results (odds ratios ORs, with 95% confidence intervals/credible intervals) of network meta-analysis for RCTs/Q-RCTs reporting global effect.
Table D2 (pg 45): Results (weighted mean difference WMD, with 95% confidence intervals/credible intervals) of network meta-analysis for RCTs/Q-RCTs reporting
SC
pain intensity.
M AN U
Results of sensitivity analyses (pg 46) (a narrative summary)
AC C
EP
TE D
Appendix E (pg 47): Assessment of model fit and between study heterogeneity
2
ACCEPTED MANUSCRIPT
WEB APPENDIX A: WINBUG CODES These are based on those published on the Bristol University network meta-analysis webpage (https://www.bris.ac.uk/cobm/research/mpes/mtc.html)
M AN U
SC
RI PT
GLOBAL #Random-effects model for multi-arm trials model{ for(i in 1:NS){ w[i,1] 1.0 favours intervention compared with control.
43
U
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intra-operative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; S intradiscal injections; T Spinal cord simulations; U Radiofrequency treatment ; N number of studies included in conventional pairwise meta-analysis.
44
ACCEPTED MANUSCRIPT
Table D2: Results (weighted mean difference WMD, with 95% confidence intervals/credible intervals) of network meta-analysis for
N=1, 6.0 (-4, 16)
N=1, -2.00 (-12, 8) N=1, 9.00 (-4, 22)
N=8,-5.17 (-12, 2) N=2,18.01 (6, 30)
D 4.56 (-18, 28) 6.49 (-5, 17) -6.74 (-30, 17) 8.64 (-13, 31) 4.06 (-31, 39) -14.55 (-44, 15) 5.39 (-19, 30) 9.93 (-10, 30) 2.80 (-15, 21) 27.64 (-5, 61) 14.43 (-8, 36) 26.83 (-10, 65) 22.64 (-11, 56) -1.79 (-23, 20) 23.39 (-3, 50)
N=1, -7.00 (-14, -0.4)
N=1, 35.00 (-25, 95) N=1, -0.63 (-15, 14)
E 1.84 (-23, 26) -11.36 (-34, 11) 4.10 (-26, 35) -0.5 (-27, 26) -19.04 (-56, 17) 0.76 (-26, 28) 5.38 (-23, 35) -1.73 (-30, 27) 23.14 (-16, 63) 9.80 (-21, 40) 22.23 (-20, 66) 18.02 (-6, 43) -6.39 (-37, 24) 18.69 (-15, 53)
F -13.23 (-38, 12) 2.19 (-20, 25) -2.39 (-38, 34) -21.0 (-51, 9) -1.12 (-28, 26) 3.41 (-17, 25) -3.69 (-23, 17) 21.22 (-12, 56) 7.97 (-12, 28) 20.3 (-17, 59) 16.11 (-18, 51) -8.28 (-29, 13) 16.86 (-10, 44)
G 15.40 (-15, 47) 10.8 (-24, 46) -7.752 (-45, 29) 12.16 (-10, 34) 16.68 (-13, 47) 9.62 (-19, 39) 34.56 (-5, 75) 21.2 (-10, 52) 33.6 (-10, 77) 29.37 (-4, 63) 4.92 (-26, 36) 30.11 (-5, 65)
-4.65 (-45, 35) -23.2 (-58, 11) -3.30 (-36, 29) 1.31 (-15, 17) -5.85 (-33, 21) 18.99 (-6, 44) 5.78 (-24, 34) 18.15 (-12, 49) 13.93 (-25, 53) -10.38 (-40, 18) 14.66 (-18, 47)
I -18.67 (-64, 26) 1.25 (-37, 39) 5.90 (-33, 46) -1.27 (-40, 38) 23.61 (-24, 72) 10.36 (-31, 50) 22.64 (-27, 74) 18.54 (-17, 55) -5.89 (-46, 34) 19.14 (-24, 62)
N=1,13.00 (2, 24)
N=1, -9.3 (-23, 4.9)
N=1, 18.00 (8, 28) N=1, 22.50 (11, 35)
N=2, 3.19 (-13, 19)
H
N=1, -26.66 (-38, -15)
N=1, -7.0 (-5, 19)
SC
C 1.65 (-20, 23) 6.23 (-14, 27) 8.12 (-16, 32) -5.09 (-14, 4) 10.3 (-19, 40) 5.75 (-28, 40) -12.81 (-49, 23) 7.04 (-13, 27) 11.61 (-17, 41) 4.47 (-23, 32) 29.48 (-9, 69) 16.11 (-14, 46) 28.5 (-14, 71) 24.26 (-8, 57) -0.16 (-30, 30) 24.99 (-9, 59)
N=1, -25.00 (-42, -8)
N= 1, 3.36 (-15, 21)
M AN U
N=1, -6.1 (-11, -0.8)
N=4, -6.26 (-15, 3)
N=1, 19.00 (8, 30)
J
19.86 (-18, 58) 24.48 (-9, 59) 17.37 (-16, 51) 42.28 (-0.1, 85) 28.89 (-6, 64) 41.4 (-4, 88) 37.07 (-6, 81) 12.74 (-22, 47) 37.93 (0.2, 76)
TE D
B -7.35 (-25, 10) -5.70 (-22, 11) -1.07 (-24, 22) 0.81 (-19, 20) -12.46 (-32, 7) 2.99 (-23, 30) -1.56 (-37, 34) -20.16 (-54, 13) -0.33 (-20, 20) 4.31 (-21, 30) -2.86 (-27, 21) 22.01 (-14, 59) 8.78 (-19, 35) 21.17 (-19, 62) 16.97 (-17, 51) -7.49 (-34, 19) 17.62 (-13, 48)
N=1, -5.40 (-24, 13)
EP
-4.61 (-22, 13) -11.96 (-34, 10) -10.3 (-19, -2) -5.68 (-28, 17) -3.82 (-14, 6) -17.04 (-41, 7) -1.64 (-21, 19) -6.26 (-41, 29) -24.8 (-53, 3) -4.95 (-30, 20) -0.36 (-18, 19) -7.49 (-25, 11) 17.35 (-14, 50) 4.19 (-17, 25) 16.51 (-19, 53) 12.26 (-21, 46) -12.11 (-32, 8) 13.03 (-12, 38)
N=7, -8.11 (-19, 3) N=2, -5.32 (-12, 1)
AC C
A
RI PT
RCTs/Q-RCTs reporting pain intensity
K
4.60 (-26, 36) -2.57 (-32, 28) 22.38 (-18, 64) 9.04 (-24, 41) 21.5 (-22, 66) 17.26 (-19, 54) -7.27 (-40, 25) 17.97 (-18 54)
L
-7.11 (-33, 18) 17.73 (-12, 48) 4.51 (-24, 32) 16.81 (-18, 51) 12.66 (-26, 50) -11.73 (-40, 15) 13.36 (-18, 44)
M 24.8 (-12, 62) 11.62 (-16, 39) 23.98 (-16, 65) 19.84 (-18, 57) -4.59 (-32, 22) 20.54 (-11, 51)
N=2, -1.09 (-7, 5)
N -13.24 (-53, 24) -0.85 (-18, 16) -5.13 (-52, 41) -29.52 (-68, 8) -4.301 (-45, 36)
N=2, -8.23 (-18, 2)
O 12.46 (-29, 55) 8.17 (-31, 48) -16.23 (-33, 1) 8.86 (-24, 42)
P -4.21 (-54, 45) -28.63 (-71, 12) -3.49 (-48, 40)
Q -24.45 (-64, 15) 0.70 (-41, 42)
R 25.09 (-7, 57)
Lower triangle includes the findings of the network meta-analysis (posterior median weighted mean differences WMDs plus 95% credible intervals) conducted in the Bayesian statistical package WinBUGS; upper triangle includes the findings of the direct standard pairwise meta-analyses (WMD plus confidence intervals) conducted using STATA Statistically significant findings have been shaded (significance assessment made on data rounded to decimal places) WMD > 0 (representing reduction in pain) favours intervention compared with control A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intra-operative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; U Radiofrequency treatment; N number of studies included in conventional pairwise meta-analysis.
45
U
ACCEPTED MANUSCRIPT
Results of sensitivity analyses
The results of the sensitivity analyses excluding observational studies and non-RCTs showed broad agreement with the main analyses. For global effect, the most notable discrepancies occurred with biological agents compared with chemonucleolysis,
RI PT
conventional care, and exercise therapy. The relative effects of epidural, chemonucleolysis, non-opioids, manipulations, and acupuncture with usual care were no longer statistically
significant in the network meta-analysis restricted to RCTs and Q-RCTs. A number of other
comparisons were also no longer statistically significant, usually with the analyses restricted to RCTs having a wider credible interval (epidural injections vs chemonucleolysis, traction,
SC
exercise therapy, and spinal cord simulation SCS; and SCS vs non-opioids, manipulation, and accupuncture). The following comparison, however, did become statistically significant
M AN U
within the restricted analyses: disc surgery was more effective than radiofrequency treatment (OR 6.68, 95% CrI: 1.19 to 39.53); intra-operative injections were more effective than both non-opioids (OR 3.16, 95% CrI: 1.09 to 9.49), passive physical therapy (OR 3.89, 95% CrI: 1.11 to 13.84), opioids (OR 5.40, 95% CrI: 1.35 to 22.40), and neuropathic pain modulators (OR 4.33, 95% CrI: 1.09 to 17.51); and manipulation was more effective radiofrequency treatment (OR 8.84, 95% CrI: 1.06 to 76.22).
TE D
For pain intensity the most notable discrepancies between the network meta-analysis with and without observational studies and non-RCTs only occurred with biological agents (vs inactive control, conventional care, disc surgery, non-opioids, intra-operative interventions, acupuncture, exercise therapy, opioids, and neuropathic painmodulators). Only the mean
EP
difference between epidural injections and inactive control (WMD -10.3 95% CrI: -18.5, 2.24) remained statistically significant in the network meta-analysis restricted to RCTs and Q-RCTs, predominantly through the inflation of confidence intervals due to the removal of
AC C
evidence.
46
ACCEPTED MANUSCRIPT
Appendix E: Assessment of model fit and between study heterogeneity MODEL FIT Table E: Residual deviance data for all network meta-analyses No. of data points 190 136 107 92
Posterior mean deviance 195 140 105 90
RI PT
Model Global effect including all studies Global effect including RCTs and Q-RCTs Pain intensity including all studies Pain intensity including RCTs and Q-RCTs
SC
HETEROGENEITY Global effect
M AN U
For the network meta-analyses of global effect that included all study designs, the median between-trial variance (tau-squared) observed in the posterior distributions was 0.38 (95% CrI: 0.22 to 0.65). However, this does not give an indication of the between study heterogeneity within each intervention comparison. This information has been derived from the standard pair-wise meta-analyses shown in the Table E1 below. There was high a level of between study heterogeneity for the following comparisons: disc surgery vs conventional care; and percutaneous discectomy vs disc surgery. Table E1: Test(s) of heterogeneity for studies included in the standard pair-wise analyses for global effect which included all study types
TE D
Degrees of freedom 4 8 2 4 13 6 0 0 6 3 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
EP
Chi-squared statistic 10.5 26.58 8.82 5.01 41.08 7.74 0.00 0.00 2.32 13.10 0.18 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.76
AC C
Treatment comparators CB DA DB EA EC FA FC FD GC GE HA HK IA JF KB KC KH LA LD LH MA ND NH NK OA OF PN
P-value
I-squared**
Tau-squared
0.038 0.001 0.012 0.286 0.000 0.258 . . 0.888 0.004 0.669 . . . . . . . . . . . . . . . 0.384
60.6% 69.9% 77.3% 20.2% 68.4% 22.5% .% .% 0.0% 77.1% 0.0% .% .% .% .% .% .% .% .% .% .% .% .% .% .% .% 0.0%
0.1806 0.7921 2.2995 0.0604 0.2582 0.0659 0.0659 0.0659 0.0000 0.7505 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
47
ACCEPTED MANUSCRIPT
I-squared**
Tau-squared
74.1% 49.0% 37.7% 0.0% 57.0% .% .% 76.3%
0.3051 0.1483 0.1814 0.0000 0.3459 0.3459 0.3459 0.5917
RI PT
Treatment Chi-squared Degrees of P-value comparators statistic freedom QC 15.43 4 0.004 QE 7.85 4 0.097 RA 1.61 1 0.205 RO 0.29 1 0.593 SE 6.97 3 0.073 TC 0.00 0 . UA 0.00 0 . Overall 396.34 94 0.000 ** I-squared: the variation in OR attributable to heterogeneity
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intraoperative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; S intradiscal injections; T Spinal cord simulations; U Radio frequency treatment.
M AN U
SC
For the network meta-analyses of global effect that included only the RCTs and quasiRCTs, the median between-trial variance (tau-squared) observed in the posterior distributions was 0.30 (95% CrI: 0.12 to 0.62). There was high a level of between study heterogeneity for the following comparisons: disc surgery vs conventional care; and percutaneous discectomy vs disc surgery (see Table E2). Table E2: Test(s) of heterogeneity for studies included in the standard pair-wise analyses for global effect which included only RCTs or quasi-RCTs
AC C
EP
TE D
Treatment Chi-squared Degrees of P-value comparators statistic freedom CB 8.96 2 0.011 DA 26.58 8 0.001 DB 0.17 1 0.680 EA 5.01 4 0.286 EC 7.26 4 0.123 FA 7.74 6 0.258 FD 0.00 0 . GC 2.32 6 0.888 HA 0.18 1 0.669 KH 0.00 0 . IA 0.00 0 . JF 0.00 0 . KB 0.00 0 . KC 0.00 0 . LA 0.00 0 . LD 0.00 0 . LH 0.00 0 . MA 0.00 0 . NH 0.44 1 0.506 OA 0.00 0 . OF 0.00 0 . PN 0.76 1 0.384 QC 7.25 1 0.007 QE 0.08 1 0.783 RA 1.61 1 0.205 RO 0.29 1 0.593 SE 5.13 2 0.077 TC 0.00 0 . UA 0.00 0 . Overall 201.29 68 0.000 ** I-squared: the variation in OR attributable to heterogeneity
I-squared**
Tau-squared
77.7% 69.9% 0.0% 20.2% 44.9% 22.5% .% 0.0% 0.0% .% .% .% .% .% .% .% .% .% 0.0% .% .% 0.0% 86.2% 0.0% 37.7% 0.0% 61.0% .% .% 66.2%
0.4647 0.7921 0.0000 0.0604 0.1757 0.0659 2.2417 0.0000 0.0000 0.3077 0.0000 0.3077 0.0000 0.0000 0.0000 2.2417 0.3077 0.0000 0.0000 0.0000 0.3077 0.0000 2.2417 0.0000 0.1814 0.0000 0.3077 2.2417 0.1814 0.4619
48
ACCEPTED MANUSCRIPT
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intraoperative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; S intradiscal injections; T Spinal cord simulations; U Radiofrequency treatment.
RI PT
Pain intensity
SC
For the network meta-analyses of pain intensity that included all study designs, the median between-trial variance (tau-squared) observed in the posterior distributions was 119.9 (95% CrI: 64.34 to 227.9). There was high a level of between study heterogeneity for the following comparisons: disc surgery vs inactive control; non-opioids vs inactive control; biological agents vs inactive control; intra-operative interventions vs disc surgery; and non-opioids vs epidural injections (see Table E3). Table E3: Test(s) of heterogeneity for studies included in the standard pair-wise analyses for pain intensity which included all study types P
I-squared**
M AN U
Treatment Heterogeneity degrees of comparator Statistic freedom (number of studies) DA 56.13 6 EA 0.00 0 FA 18.77 3 HA 0.00 0 JA 0.00 0 LA 0.00 0 MA 13.69 1 RA 0.00 0 UA 0.00 0 CB 1.02 1 DB 0.03 1 KB 0.00 0 EC 6.20 2 FC 0.00 0 GC 29.82 7 KC 0.00 0 FD 7.92 1 JD 0.00 0 LD 0.00 0 IE 0.00 0 QE 0.00 0 MF 0.00 0 OF 0.00 0 LH 3.86 1 NH 0.00 0 PN 0.16 1 RO 0.67 1 ** I-squared: the variation in WMD attributable to heterogeneity
AC C
EP
TE D
0.000 . 0.000 . . . 0.000 . . 0.313 0.868 . 0.045 . 0.000 . 0.005 . . . . . . 0.049 . 0.689 0.414
89.3% .% 84.0% .% .% .% 92.7% .% .% 1.8% 0.0% .% 67.7% .% 76.5% .% 87.4% .% .% .% .% .% .% 74.1% .% 0.0% 0.0%
Tau-squared
198.6672 0.0000 65.8051 0.0000 0.0000 0.0000 535.7751 0.0000 0.0000 0.7251 0.0000 0.0000 56.4674 0.0000 72.3876 0.0000 62.9147 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 98.4535 0.0000 0.0000 0.0000
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intraoperative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; U Radiofrequency treatment.
For the network meta-analyses of pain intensity that included only the RCTs and quasiRCTs, the median between-trial variance (tau-squared) observed in the posterior distributions was 122.4 (95% CrI: 62.15 to 249.9). There was high a level of between study heterogeneity for the following comparisons: disc surgery vs inactive control; non-opioids vs inactive control; biological agents vs inactive control; intra-operative interventions vs disc surgery; and non-opioids vs epidural injections (see Table E4).
49
ACCEPTED MANUSCRIPT
Table E4: Test(s) of heterogeneity for studies included in the standard pair-wise analyses for pain intensity which included only RCTs or quasi-RCTs I-squared**
Tau-squared
0.000 . 0.000 . . . 0.000 . . . 0.868 . . 0.000 . 0.005 . . . . 0.049 . 0.689 0.414
89.3% .% 84.0% .% .% .% 92.7% .% .% .% 0.0% .% .% 76.5% .% 87.4% .% .% .% .% 74.1% .% 0.0% 0.0%
198.6672 0.0000 65.8051 0.0000 0.0000 0.0000 535.7751 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 72.3876 0.0000 62.9147 0.0000 0.0000 0.0000 0.0000 98.4535 0.0000 0.0000 0.0000
SC
RI PT
P
M AN U
Treatment Heterogeneity degrees of comparator Statistic freedom (number of studies) DA 56.13 6 EA 0.00 0 FA 18.77 3 HA 0.00 0 JA 0.00 0 LA 0.00 0 MA 13.69 1 RA 0.00 0 UA 0.00 0 CB 0.00 0 DB 0.03 1 KB 0.00 0 EC 0.00 0 GC 29.82 7 KC 0.00 0 FD 7.92 1 LD 0.00 0 IE 0.00 0 QE 0.00 0 OF 0.00 0 LH 3.86 1 NH 0.00 0 PN 0.16 1 RO 0.67 1 ** I-squared: the variation in WMD attributable to heterogeneity
AC C
EP
TE D
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intraoperative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; U Radiofrequency treatment.
50
ACCEPTED MANUSCRIPT
Table 1: Treatment categorisation
Treatment strategy
Treatment strategy Codea
Inactive control
A
Type of treatment Placebo in any type or format – tablet, injection, epidural etc.
RI PT
a
Sham treatment in any format No treatment B
Conservative therapy
SC
Conventional care
Conventional care
Non-surgical treatments
Disc surgery
C
M AN U
General practitioner care Discectomy Micro-discectomy Laminectomy Hemilaminectomy
TE D
Surgical decompression
D
AC C
EP
Epidural injections (includes spinal nerve block)
Chemonucleolysis
Non-opioids
Lumbar epidural injection Transforminal epidural injection Intraforaminal injection Interlaminar epidural injection Caudal epidural injection Periradicular injection Spinal nerve root block
E
Chymopapain Collagenase Ozone
F
Conventional pain or anti-inflammatory medication used as oral, IV or intramuscular: Steroids
ACCEPTED MANUSCRIPT
Treatment a strategy Code
a
Treatment strategy
Type of treatment Non-steroidal anti-inflammatory drugs (NSAIDs) Paracetamol
G
Barrier membranes:
RI PT
Intra-operative interventions
-
Anti-adhesion barrier
-
Fat graft
Intra-operative steroid +/- LA H
Mechanical traction
SC
Traction
Antigravitational traction
M AN U
Auto-traction
Manual traction
Manipulation
I
Chiropractic Osteopathic
J
Exercise therapy
K
L
AC C
EP
Passive physical therapy
TE D
Acupuncture
Biological agents
Acupuncture Exercise therapy Isometric exercises Mobilising and strengthening exercises Ultrasound Transcutaneous electrical nerve stimulation (TENS) Infra-red heat Physical therapy programme (hot pack, continuous ultrasound, and diadynamic currents) Conservative physiotherapy
M
Cytokine modulating treatments targeting tumour necrosis factor alpha: - Entanercept - Infliximab - Autologous Conditioned Serum
ACCEPTED MANUSCRIPT
Treatment strategy
Treatment a strategy Code
Bed rest
N
Opioids
O
Oral, IV or intramuscular opioids
Education/Advice
P
Advice to keep active
Type of treatment
RI PT
a
Advised to continue activities of daily living Percutaneous discectomy
Q
Automated percutaneous discectomy
SC
Percutaneous automated nucleotomy Nucleoplasty
Neuropathic painmodulators
R
M AN U
Lasar discectomy
Pharmaceutical treatment used for neuropathic pain: Anti-epileptic medication Tricyclic antidepressants
S
Spinal cord stimulation
T
Radiofrequency treatment a
TE D
Intradiscal injections
U
Interventions are summarised using these codes for displaying the results of the network-meta-analyses
AC C
EP
IV intra-venous
ACCEPTED MANUSCRIPT
D 0.45 (0.2, 0.9) 0.63 (0.3, 1.2) 1.05 (0.4, 2.4) 0.37 (0.2, 0.8) 1.4 (0.3, 7.0) 2.27 (0.3, 19.9) 0.31 (0.1, 0.9) 0.43 (0.2, 1.1) 4.8 (0.2, 278) 0.39 (0.1, 1.3) 0.33 (0.1, 1.1) 0.50 (0.1, 2.5) 0.35 (0.2, 0.8) 0.4 (0.1, 1.3) 0.2 (0.1, 0.6) 0.80 (0.1, 5.1) 0.16 (0.03, 0.9)
N=1, 6.71 (0.8, 58.8) N=1, 0.45 (0.2, 1.4)
N=1, 4.27 (1.5, 12.4)
N=7, 1.49 (1, 2.2)
N=1, 0.2 (0.1, 0.6)
N=1, 3.27 (0.8, 13.9)
F 1.67 (0.7, 4.1) 0.59 (0.2, 1.5) 2.24 (0.4, 11.4) 3.62 (0.5, 28.4) 0.5 (0.2, 1.5) 0.68 (0.2, 2) 7.71 (0.3, 449) 0.62 (0.2, 2.4) 0.53 (0.2, 1.6) 0.80 (0.1, 4.4) 0.56 (0.2, 1.4) 0.65 (0.2, 2) 0.32 (0.1, 0.97) 1.30 (0.2, 8.4) 0.25 (0.04, 1.5)
N=1, 10 (0.7,166.7)
N=1, 1.46 (0.7, 2) N=1, 0.77 (0.18, 3.2)
N=4, 2.03 (0.8, 5.4)
E 1.39 (0.7, 3.0) 2.32 (1.4, 4.0) 0.8 (0.3, 1.9) 3.1 (0.6, 16.3) 5.0 (0.7, 45.5) 0.7 (0.3, 1.9) 0.95 (0.3, 2.8) 10.77 (0.5, 625) 0.86 (0.2, 3.2) 0.73 (0.2, 2.5) 1.12 (0.2, 6.0) 0.78 (0.5, 1.3) 0.89 (0.3, 3.0) 0.44 (0.2, 1) 1.80 (0.3, 10.2) 0.35 (0.1, 2.1)
N=1, 4.72 (1.9, 11.4)
G 0.35 (0.1, 0.9) 1.34 (0.2, 7.5) 2.18 (0.3, 20.7) 0.30 (0.1, 0.9) 0.41 (0.1, 1.3) 4.65 (0.2, 277) 0.37 (0.1, 1.5) 0.32 (0.1, 1.2) 0.48 (0.1, 2.7) 0.34 (0.2, 0.7) 0.39 (0.1, 1.4) 0.19 (0.1, 0.5) 0.77 (0.1, 4.6) 0.15 (0.02, 0.96)
N=1, 0.87 (0.3, 2.7)
H 3.83 (0.7, 21.4) 6.21 (0.8, 59.0) 0.86 (0.3, 2.4) 1.17 (0.4, 3.2) 13.3 (0.5, 792) 1.06 (0.3, 3.8) 0.9 (0.2, 3.4) 1.37 (0.3, 7.2) 0.96 (0.4, 2.5) 1.1 (0.3, 4.1) 0.54 (0.2, 1.8) 2.18 (0.3, 14.8) 0.43 (0.1, 2.7)
I 1.63 (0.1, 22.3) 0.22 (0.04, 1.4) 0.3 (0.05, 1.8) 3.48 (0.1, 252) 0.28 (0.04, 2.0) 0.24 (0.04, 1.5) 0.36 (0.04, 3.4) 0.25 (0.04, 1.4) 0.29 (0.04, 1.8) 0.14 (0.02, 0.9 0.57 (0.1, 6.06) 0.11 (0.00, 1.1)
J 0.14 (0.01, 1.2) 0.19 (0.02, 1.7) 2.15 (0.05, 185) 0.17 (0.01, 1.7) 0.14 (0.01, 1.3) 0.22 (0.02, 2.8) 0.15 (0.02, 1.2) 0.18 (0.02, 1.6) 0.09 (0.01, 0.8) 0.35 (0.02, 5.1) 0.07 (0.01, 0.9)
N=1, 1.37 (0.5,3.8)
N=1, 0.93 (0.5, 1.9)
N=1, 0.57 (0.2, 1.7)
N=5, 0.67 (0.4, 1.3)
N=1, 1.13 (0.4, 3.6)
N=5, 0.65 (0.4, 1.1)
N=4, 0.46 (0.2, 1)
N=1, 0.18 (0.1, 0.7)
N=1, 1 (0.1, 7.1)
N=1, 2.2 (0.6, 7.6)
K
1.37 (0.4, 5.0) 15.63 (0.6, 977) 1.24 (0.4, 4.3) 1.05 (0.2, 4.7) 1.6 (0.3, 8.3) 1.13 (0.4, 3.3) 1.29 (0.3, 5.6) 0.64 (0.2, 2.3) 2.56 (0.4, 18.5 0.51 (0.1, 3.6)
N=2, 2.01 (0.8, 5.2)
N=1, 0.22 (0.1, 0.9)
SC
N=14, 0.5 (0.4, 0.7)
C 1.42 (0.7, 2.9) 0.64 (0.5, 0.9) 0.89 (0.4, 1.9) 1.48 (0.9, 2.5) 0.52 (0.2, 1.2) 1.98 (0.4, 10.5) 3.22 (0.4, 29.2) 0.44 (0.2, 1.2) 0.61 (0.2, 1.8) 6.89 (0.3, 400) 0.55 (0.2, 2) 0.47 (0.1, 1.6) 0.71 (0.1, 3.8) 0.5 (0.3, 0.8) 0.57 (0.17, 1.9) 0.28 (0.1, 0.7) 1.13 (0.2, 6.3) 0.23 (0.04, 1.3)
N=2, 1.11 (0.6, 2) N=1, 1.53 (0.6, 4.2)
M AN U
3.0 (1.9, 4.9) 4.26 (2.1, 8.8) 1.92 (1.1, 3.4) 2.66 (1.2, 6.2) 4.5 (2.2, 9.0) 1.56 (0.7, 3.7) 5.96 (1.1, 32.2) 9.72 (1.2, 89.2) 1.34 (0.5, 3.4) 1.82 (0.6, 5.5) 20.75 (0.9, 1213) 1.65 (0.5, 6.0) 1.41 (0.4, 5.1) 2.14 (0.4, 11.3) 1.50 (0.7, 3.0) 1.72 (0.5, 6.0) 0.85 (0.3, 2.3) 3.41 (0.6, 20.3) 0.68 (0.1, 4.1)
N=7, 2.7 (1.7, 4.3)
N=7, 1.68 (1.1, 2.5)
TE D
B
N=5, 2.56 (1.6, 4.1)
EP
0.82 (0.4, 1.6) 2.46 (1.3, 4.5) 3.48 (2.1,5.8) 1.57 (0.9, 2.8) 2.18 (1.3, 3.8) 3.64 (1.7, 7.8) 1.28 (0.6, 2.7) 4.88 (1.1, 23) 7.92 (1.1, 67) 1.09 (0.4, 3) 1.49 (0.6, 3.9) 16.83 (0.8, 947) 1.35 (0.4, 4.7) 1.15 (0.4, 3.4) 1.75 (0.3, 9.0) 1.23 (0.6, 2.6) 1.41 (0.5, 4.0) 0.7 (0.3, 1.9) 2.79 (0.4, 17.2) 0.56 (0.1, 2.9)
N=9, 2.58 (1.3, 5.3) N=3, 5.46 (0.8, 38.5)
AC C
A
RI PT
Table 2: Results (odds ratios, with 95% confidence intervals/credible intervals) of the network meta-analysis for global effect
L
11.41 (0.4, 699) 0.9 (0.2, 4.0) 0.77 (0.2, 3.3) 1.17 (0.2, 7.1) 0.82 (0.3, 2.6) 0.94 (0.2, 3.8) 0.47 (0.1, 1.8) 1.87 (0.2, 14.1) 0.37 (0.1, 2.5)
M 0.08 (0.0, 2.3) 0.07 (0.0, 1.8) 0.1 (0.0, 3.5) 0.07 (0.0, 1.7) 0.08 (0.0, 2.2) 0.04 (0.0, 1.1) 0.16 (0.0, 6.0) 0.03 (0.0, 1.1)
N=2, 1.31 (0.8, 2.3)
N 0.85 (0.2, 4.5) 1.30 (0.5, 3.7) 0.91 (0.2, 3.6) 1.04 (0.2, 5.4) 0.51 (0.1, 2.4) 2.07 (0.2, 17.52) 0.41 (0.05, 3.4)
N=2, 0.78 (0.4, 1.7)
O 1.52 (0.2, 10.8) 1.07 (0.3, 4) 1.23 (0.4, 3.6) 0.6 (0.1, 2.7) 2.42 (0.3, 20.3) 0.48 (0.1, 3.5)
P 0.70 (0.1, 3.9) 0.8 (0.1, 5.6) 0.4 (0.1, 2.6) 1.6 (0.1, 18) 0.32 (0.03, 3.3)
Q 1.15 (0.3, 4.2) 0.57 (0.2, 1.5) 2.27 (0.4, 13.8) 0.45 (0.1, 2.8)
R 0.50 (0.1, 2.1) 2.0 (0.2, 16.3) 0.40 (0.1, 2.9)
S 4.0 (0.6, 28) 0.80 (0.1, 5.7)
T 0.20 (0.02, 2.3)
Lower triangle includes the findings of the network meta-analysis (posterior median odds ratios ORs plus 95% credible intervals) conducted in the Bayesian statistical package WinBUGS; upper triangle includes the findings of the direct standard pairwise meta-analyses (OR plus confidence intervals) conducted using STATA Statistically significant findings have been shaded (significance assessment made on data rounded to decimal places)
1
U
ACCEPTED MANUSCRIPT
OR > 1.0 favours intervention compared with control. A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intra-operative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L
RI PT
Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; S intradiscal injections; T Spinal cord simulations; U
AC C
EP
TE D
M AN U
SC
Radiofrequency treatment; N number of studies included in conventional pairwise meta-analysis.
2
ACCEPTED MANUSCRIPT
Table 3: Results (weighted mean difference, with 95% confidence intervals/credible intervals) of the network meta-analysis for pain
-1.89 (-17, 14) 3.02 (-10, 16) 7.02 (-9, 23) -5.12 (-14, 4) 7.79 (-17, 33) 2.4 (-27, 32) -5.21 (-29, 18) 6.95 (-12, 26) 9.14 (-15, 34) -9.96 (-30, 10) 26.7 (-8, 62) 14.51 (-10, 40) 25.83 (-13, 65) 21.14 (-7, 49) -2.07 (-27, 23) 22.53 (-7, 52)
D 4.85 (-13, 23) 8.86 (-1, 19) -3.27 (-21, 15) 9.61 (-11, 31) 4.19 (-28, 36) -3.42 (-22, 15) 8.79 (-13, 31) 10.96 (-8, 31) -8.10 (-22, 6) 28.56 (-3, 61) 16.32 (-5, 38) 27.61 (-8, 64) 22.94 (-8, 53) -0.21 (-21, 21) 24.38 (-2, 51)
N=1, 1.60 (-8, 11) N=2,18.01 (6, 30)
N=8,-5.17 (-12, 2) N=1, 4.00 (-6, 14) N=1, -0.63 (-15, 14)
F -12.09 (-30, 6) 0.69 (-21, 23) -4.7 (-37, 27) -12.28 (-33, 8) -0.07 (-23, 23) 2.10 (-18, 23) -16.99 (-32, -3) 19.65 (-13, 53) 7.44 (-13, 27) 18.77 (-17, 56) 14.14 (-17, 45) -9.01 (-30, 12) 15.52 (-11, 42)
G 12.89 (-14, 40) 7.48 (-23, 38) -0.13 (-26, 25) 12.07 (-9, 33) 14.26 (-11, 40) -4.85 (-27, 17) 31.84 (-4, 68) 19.61 (-7, 46) 30.99 (-9, 71) 26.2 (-3, 55) 3.06 (-23, 30) 27.64 (-3, 58)
N=1, -7.00 (-14, -0.4)
N=2, -9.91 (-43, 23)
N=1, 35.00 (-25, 95)
N=1, -9.3 (-23, 4.9)
N=1, -2.00 (-12, 8) N=1, 9.00 (-4, 22)
E 4.04 (-15, 22) -8.11 (-24, 8) 4.79 (-22, 32) -0.63 (-27, 26) -8.25 (-34, 17) 3.95 (-19, 27) 6.15 (-19, 32) -12.95 (-35, 9) 23.65 (-12, 60) 11.47 (-15, 38) 22.8 (-17, 63) 18.14 (-6, 42) -5.07 (-31, 21) 19.52 (-11, 50)
N=1, -25.00 (-42, -8)
SC
N=3, 2.36 (-8, 13)
C
N= 1, 3.36 (-15, 21)
I -7.58 (-44, 29) 4.58 (-30, 39) 6.81 (-30, 44) -12.3 (-46, 22) 24.38 (-20, 69) 12.15 (-25, 49) 23.39 (-24, 71) 18.8 (-17, 55) -4.37 (-42, 33) 20.19 (-20, 61)
N=1, -26.66 (-38, -15)
N=1,13.00 (2, 24)
N=1, 18.00 (8, 28) N=1, -40.50 (-58, -23)
N=2, 3.19 (-13, 19)
H -5.42 (-43, 32) -13.02 (-41, 14) -0.75 (-31, 29) 1.4 (-14, 18) -17.72 (-42, 6) 19.02 (-6, 43) 6.74 (-22, 35) 18.11 (-12, 48) 13.38 (-23, 49) -9.82 (-38, 18) 14.84 (-17, 46)
M AN U
-7.13 (-21, 6) -8.96 (-23, 6) -4.17 (-22, 14) -0.10 (-16, 16) -12.21 (-29, 4) 0.68 (-24, 26) -4.77 (-37, 27) -12.35 (-36, 11) -0.19 (-19, 19) 1.98 (-21, 26) -17.03 (-37, 2) 19.62 (-15, 55) 7.41 (-18, 32) 18.71 (-20, 57) 13.97 (-16, 44) -9.13 (-34, 16) 15.36 (-14, 45)
N=2, -7.0 (-12, -2)
N=4, -6.26 (-15, 3)
N=1, 22.50 (10, 35)
N=1, 19.00 (8, 30)
J
12.19 (-16, 41) 14.45 (-12, 41) -4.71 (-27, 18) 32.01 (-4, 69) 19.75 (-8, 47) 31.07 (-9, 71) 26.36 (-9, 62) 3.21 (-24, 31) 27.76 (-3, 59)
TE D
B
N=1, -5.40 (-24, 13)
EP
-2.44 (-18, 13) -9.54 (-25, 6) -11.4 (-19, -4) -6.62 (-25, 11) -2.54 (-12, 6) -14.64 (-33, 3) -1.81 (-22, 18) -7.24 (-39, 25) -14.82 (-34, 4) -2.62 (-25, 20) -0.46 (-18, 18) -19.51 (-33, -6) 17.13 (-14, 49) 4.92 (-16, 25) 16.22 (-19, 52) 11.6 (-19, 42) -11.6 (-32, 8) 12.98 (-12, 38)
N=7, -8.11 (-19, 3) N=2, -5.32 (-12, 1)
AC C
A
RI PT
intensity
K
2.16 (-26, 31) -16.93 (-43, 8) 19.77 (-18, 58) 7.58 (-22, 37) 18.9 (-23, 61) 14.17 (-19, 47) -8.99 (-39, 21) 15.55 (-18, 49)
L
-19.09 (-42, 3) 17.62 (-12, 47) 5.42 (-23, 32) 16.69 (-17, 50) 11.96 (-24, 47) -11.13 (-39, 16) 13.42 (-18, 44)
M 36.68 (3, 71) 24.45 (0.78, 48) 35.76 (-2, 74) 31.04 (-2, 64) 7.93 (-16, 32) 32.48 (5, 61)
N=2, -1.09 (-7, 5)
N -12.19 (-50, 25) -0.90 (-18, 16) -5.46 (-49, 37) -28.78 (-66, 8) -4.16 (-44, 35)
N=2, -8.23 (-18, 2)
O 11.27 (-29, 53) 6.55 (-29 43) -16.52 (-33, 0.26) 8.01 (-24, 40)
P -4.62 (-52, 41) -27.81 (-69, 12) -3.28 (-46, 40)
Q -23.11 (-59 13) 1.4 (-38, 41)
R 24.55 (-7, 57)
Lower triangle includes the findings of the network meta-analysis (posterior median weighted mean differences WMDs plus 95% credible intervals) conducted in the Bayesian statistical package WinBUGS; upper triangle includes the findings of the direct standard pairwise meta-analyses (WMD plus confidence intervals) conducted using STATA Statistically significant findings have been shaded (significance assessment made on data rounded to decimal places) WMD > 0 (representing reduction in pain) favours intervention compared with control
1
U
ACCEPTED MANUSCRIPT
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intra-operative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; U Radiofrequency treatment; N number of studies
AC C
EP
TE D
M AN U
SC
RI PT
included in conventional pairwise meta-analysis.
2
ACCEPTED MANUSCRIPT
Figure 1: flow diagram showing the number of references identified, publications retrieved for assessment, and studies included in the review
References identified by electronic searches after de-duplication N=33560 References obtained from other sources (searching bibliographies, reference lists of reviews) N=30
RI PT
References rejected on the basis of title and/or abstract N=32812
Ongoing/completed studies with no available outcome data N=42 (Most identified through trial registries)
AC C
EP
TE D
Studies included in the HTA review* N=270 (Publications N=388)
Studies that reported data on back specific functional status but did not report useable data on global effect or pain intensity N=5
Studies included in analysis of global effect N=95 (68 RCTs/Q-RCTs)
Publications excluded from HTA review* because they did not meet criteria relating to design, population, intervention or outcome N=241
M AN U
Publications that could not be assessed for inclusion because: Unable to translate N=93 Unavailable from interlibrary loans N=14
SC
Publications retrieved in full for detailed evaluation and assessment for inclusion N=777
Studies included in refined network meta-analyses N=122 (90 RCTs/Q-RCTs)
Studies excluded from initial network metaanalyses because they evaluated mixed treatments, postoperative interventions or interventions classified under the same treatment strategy N=142 Studies excluded from refined network metaanalyses: experimental intervention no longer part of the new nonopioid category N=3
Studies included in analysis of pain intensity N=53 (46 RCTs/Q-RCTs)
*The current study represents further refinements of initial network meta-analyses conducted as part of a broader Health Technology Assessment (HTA) review evaluating the clinical and cost effectiveness of sciatica treatments;16 the original review included a narrative synthesis, conventional pairwise analyses, network meta-analyses, a review of economic evaluations and an economic evaluation. Five studies reported data on back specific functional status (an outcome considered as part of the initial network meta-analyses), but did not report useable data for global effect or pain intensity.
ACCEPTED MANUSCRIPT
Figure 2: Network of treatment strategies for sciatica for comparative studies reporting global effect
RI PT
Each treatment strategy is a node in the network. The links between the nodes are arms in head-to-head (direct) comparisons in eligible studies. The numbers along the link lines indicate the number of studies or pairs of study arms for that link in the network, with the observed I2 statistic (based on the data from the pairwise meta-analyses) is presented in brackets (0-40% might not be important; 30-60% moderate heterogeneity; 50-90% substantial heterogeneity 75% to 100% considerable heterogeneity34). Links that do not have any randomised controlled trial (RCT) or Quasi-
AC C
EP
TE D
M AN U
SC
RCT evidence are indicated in green.
Biological agents (M)
Manipulation (I) 1
Radiofrequency treatment (U)
Inactive control (A)
1
1
1
9 (70%)
1
Epidural injections / Nerve block (D) 3 (77%) 1
TE D
1 2 (0%)
Passive physical therapy (L)
1
1
Exercise therapy (K)
Education / Advice (P)
2 (0%)
2 (0%) Opioids (O)
Conventional care (B) 7 (79%) Disc surgery (C) 7 (0%)
1
1 5 (20%)
AC C
1
EP
1 Traction (H)
SC 1
M AN U
Acupuncture (J)
Neuropathic painmodulators (R)
2 (38%)
7 (23%)
Non-opioids (F)
1
1
RI PT
ACCEPTED MANUSCRIPT
14 (68%)
5 (74%) Percutaneous discectomy (Q)
1
Spinal cord stimulation (T)
5 (49%)
1
1
Bed rest (N)
4 (77%) Intra-operative interventions (G)
Chemonucleolysis (E) 4 (57%) Intradiscal injections (S)
ACCEPTED MANUSCRIPT
Figure 3: Network of treatment strategies for sciatica for comparative studies reporting pain intensity
RI PT
Each treatment strategy is a node in the network. The links between the nodes are arms in head-to-head (direct) comparisons in eligible studies. The numbers along the link lines indicate the number of studies or pairs of study arms for that link in the network, with the observed I2 statistic (based on the data from the pairwise meta-analyses) is presented in brackets (0-40% might not be important; 30-60% moderate heterogeneity; 50-90% substantial heterogeneity 75% to 100% considerable heterogeneity34). Links that do not have any randomised controlled trial (RCT) or Quasi-
AC C
EP
TE D
M AN U
SC
RCT evidence are indicated in green.
ACCEPTED MANUSCRIPT
RI PT
Radiofrequency treatment (U)
1
Inactive control (A)
Opioids (O)
8 (77%)
Conventional care (B)
1
1
Non-opioids (F)
2 (0%)
Epidural injections / Nerve block (D)
1
EP
1
TE D
7 (89%)
2 (87%)
Passive Physical Therapy (L)
2 (2%) 1 Disc surgery (C)
1
Percutaneous discectomy (Q)
3 (68%)
1
1
Chemonucleolysis (E)
1
1 1
AC C
1
1
2 (74%)
Traction (H) Education / Advice (P)
Biological agents (M)
1
4 (84%)
Intra-operative interventions (G)
SC
2 (0%)
1
M AN U
Neuropathic painmodulators (R)
2 (93%)
2 (0%)
Acupuncture (J)
1 Bed rest (N)
Exercise therapy (K)
Manipulation (I)
ACCEPTED MANUSCRIPT
Figure 4: Plot of the odds ratios (ORs) of global effects for different treatment strategies compared with inactive control from the network meta-analysis NB The data has been spun around so that effect estimates that favour the intervention are shown on the right hand side. This means that the an OR 0 represents a reduction in pain intensity in favour of the intervention
Treatment
ES (95% CI)
Biological agents (M)
-19.58 (-32.69, -6.47)
Acupuncture (J)
-14.87 (-33.60, 3.86)
SC
category
Intra-operative interventions (G)
-14.71 (-32.66, 3.24) -11.62 (-31.49, 8.25)
Neuropathic painmodulators (R)
-11.43 (-19.12, -3.74)
Epidural injections (D)
M AN U
Disc surgery (C) Manipulation (I) Chemonucleolysis (E) Exercise therapy (K) Non-opioids (F) Conventional care (B) Traction (H) Passive physical therapy (L) Opioids (O) Percutaneous discectomy (Q)
Education/Advice (P) Bed rest (N)
TE D
Radiofrequency treatment (U)
-9.60 (-25.25, 6.04) -7.22 (-38.90, 24.45) -6.59 (-24.35, 11.17) -2.63 (-24.97, 19.72) -2.58 (-11.61, 6.44) -2.45 (-17.78, 12.88) -1.77 (-21.55, 18.00) -0.35 (-18.50, 17.80) 4.89 (-15.43, 25.22) 11.50 (-18.57, 41.57) 12.96 (-11.74, 37.66) 16.32 (-18.76, 51.40) 17.22 (-13.77, 48.21)
NOTE: Weights are from random effects analysis
-40
-30
-20
-10 -5 0 5 10
20
30
40
favours inactive control
AC C
EP
favours intervention
1
S1529-9430(13)01497-6
DOI:
10.1016/j.spinee.2013.08.049
Reference:
SPINEE 55546
To appear in:
The Spine Journal
Received Date: 10 November 2011 Revised Date:
9 July 2013
Accepted Date: 23 August 2013
Please cite this article as: Lewis R, Williams NH, Sutton AJ, Burton K, Din N, Matar HE, Hendry M, Phillips CJ, Nafees S, Fitzsimmons D, Rickard I, Wilkinson C, Comparative Clinical Effectiveness of Management Strategies for Sciatica: Systematic Review and Network Meta-Analyses, The Spine Journal (2013), doi: 10.1016/j.spinee.2013.08.049. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
COMPARATIVE CLINICAL EFFECTIVENESS OF MANAGEMENT STRATEGIES FOR SCIATICA: SYSTEMATIC REVIEW AND NETWORK META-ANALYSES Ruth Lewis1, Nefyn H Williams1,2, Alex J Sutton3, Kim Burton4, NafeesUd Din1, Hosam E Matar5, Maggie Hendry1, Ceri J Phillips6, Sadia Nafees1, Deborah Fitzsimmons4, I Rickard7,
1
RI PT
Clare Wilkinson1
North Wales Centre for Primary Care Research, College of Health & Behavioural Sciences,
Bangor University, Wrexham, UK
North Wales Organisation for Randomised Trials in Health (NWORTH), Bangor University,
SC
2
Bangor, UK
Department of Health Sciences, University of Leicester, Leicester, UK
4
Spinal Research Institute, University of Huddersfield, Huddersfield, UK
M AN U
3
5
Sheffield Teaching Hospitals NHS Foundation Trust, Northern General Hospital, Sheffield, UK School of Human and Health Sciences, Swansea University, Swansea, UK
7
Betws-y-Coed, UK
Ruth Lewis - Lecturer
TE D
6
Nefyn H Williams – Senior clinical lecturer Alex J Sutton – Professor
EP
Kim Burton - Professor
NafeesUd Din – Research Associate
AC C
Hosam E Matar – Foundation II doctor Maggie Hendry - Research Fellow Ceri J Phillips - Professor SadiaNafees – Research Assistant Deborah Fitzsimmons - Senior Lecturer Ian Rickard - Patient representative Clare Wilkinson - Professor
Maggie Hendry; Ceri J Phillips; Sadia Nafees; Deborah Fitzsimmons; Ian Rickard; Clare Wilkinson
ACCEPTED MANUSCRIPT
Corresponding author:
RI PT
Ruth Lewis North Wales Centre for Primary Care Research North Wales Clinical School College of Health & Behavioural Sciences Bangor University Gwenfro Units 6-7 Wrexham Technology Park Wrexham LL13 7YP
SC
Tel +44 (0)1978 727340 Fax +44 (0)1978 311419 e-mail: [email protected]
Acknowledgement: We would like to thank Barbara France and Annie Hendry who
M AN U
helped with producing the Figures and Tables; Di Pasterfield for proof reading the manuscript; and Ian Braithwaite (Consultant Surgeon) and Rob Chakraverty (Sports Physician) for providing clinical input at various stages of the review.We would also like to thank the associated editor and the anonymous reviewers for their careful and constructive criticism which helped improve and simplify a previous version of this
TE D
paper.
This study was funded by the National Institute for Health Research (NIHR) HTA programme (project number: 06/79/01) and is published in full in Health Technology Assessment; Vol.15. The funding body had no role in design and conduct of the study; data collection; management, analyses, and interpretation of the data;
EP
preparation of the manuscript; or the decision to submit the article for publication. The views and opinions expressed therein are those of the authors and do not
AC C
necessarily reflect those of the HTA programme, NIHR, NHS or the Department of Health.
ACCEPTED MANUSCRIPT
1
COMPARATIVE CLINICAL EFFECTIVENESS OF MANAGEMENT STRATEGIES FOR
2
SCIATICA: SYSTEMATIC REVIEW AND NETWORK META-ANALYSES
3 4 ABSTRACT
6
Background: There are numerous treatment approaches for sciatica. Previous systematic
7
reviews have not compared all these strategies together.
8
Purpose: To compare the clinical effectiveness of different treatment strategies for sciatica
9
simultaneously.
SC
RI PT
5
Study design: Systematic review and network meta-analysis.
11
Methods: We searched 28 electronic databases and online trial registries, along with
12
bibliographies of previous reviews for comparative studies evaluating any intervention to
13
treat sciatica in adults, with outcome data on global effect or pain intensity. Network meta-
14
analysis methods were used to simultaneously compare all treatment strategies and allow
15
indirect comparisons of treatments between studies. The study was funded by the UK
16
National Institute for Health Research (NIHR) HTA programme; there are nopotential conflict
17
of interests.
18
Results: We identified 122 relevant studies; 90 were randomised controlled trials (RCTs) or
19
quasi-RCTs. Interventions were grouped into 21 treatment strategies. Internal and external
20
validity of included studies was very low. For overall recovery as the outcome, compared
21
with inactive control or conventional care, there was a statistically significant improvement
22
following disc surgery, epidural injections, non-opioid analgesia, manipulation, and
23
acupuncture. Traction, percutaneous discectomy and exercise therapy were significantly
24
inferior to epidural injections or surgery. For pain as the outcome, epidural injections, and
25
biological agents were significantly better than inactive control, but similar findings for disc
26
surgery were not statistically significant. Biological agents were significantly better for pain
27
reduction than bed rest, non-opioids, and opioids. Opioids, education/advice alone, bed rest,
AC C
EP
TE D
M AN U
10
1
ACCEPTED MANUSCRIPT
and percutaneous discectomy were inferior to most other treatment strategies; although
2
these findings represented large effects, they were statistically equivocal.
3
Conclusions: For the first time many different treatment strategies for sciatica have been
4
compared in the same systematic review and meta-analysis. This approach has provided
5
new data to assist shared decision-making. The findings support the effectiveness of non-
6
opioid medication, epidural injections and disc surgery. They also suggest that spinal
7
manipulation, acupuncture, and experimental treatments such as anti-inflammatory
8
biological agents, may be considered. The findings do not provide support for the
9
effectiveness of opioid analgesia, bed rest, exercise therapy, education/advice (when used
SC
RI PT
1
alone), percutaneous discectomy or traction. The issue of how best to estimate the
11
effectiveness of treatment approaches according to their order within a sequential treatment
12
pathway remains an important challenge.
13 14
M AN U
10
INTRODUCTION
16
Sciatica is the term used for the syndrome characterised by radicular leg pain, with or
17
without sensory deficits, radiating along the distribution of the sciatic nerve.1-3 In about 90%
18
of cases, it is caused by an intervertebral disc herniation resulting in nerve root irritation.4-6 It
19
is a common reason for seeking medical advice,7,8 and has considerable economic
20
consequence in terms of healthcare resources and lost productivity.7 The diagnosis and
21
management of sciatica varies considerably within and between countries,4 which may
22
reflect treatment availability, clinician preference and socio-economic variables rather than
23
evidence-based practice.
EP
AC C
24
TE D
15
25
Previous systematic reviews (including meta-analyses) have evaluated the effectiveness of
26
various individual treatment approaches for sciatica, including conservative treatments,9-12
27
epidural steroid injections,9,11,13,14 and surgical procedures.15 However, numerous treatments
28
have not been directly compared. Furthermore, in order to choose the optimal treatment(s), it
2
ACCEPTED MANUSCRIPT
would be more helpful if all candidate treatments could be compared in the same analysis,
2
as opposed to using a series of simple but inefficient standard pairwise meta-analyses
3
comparing only two treatments at a time. It has been acknowledged that there is difficulty in
4
interpreting the findings of multiple comparisons with low power, due to the small number of
5
participants or events, which are inclined to result in statistically insignificant findings.16,17
RI PT
1
6
A network meta-analysis,18 by contrast, enables the simultaneous comparison of more than
8
two treatment approaches, whilst combining data derived from both direct within-study
9
comparisons between two treatment strategies (e.g A vs B) and comparisons constructed
10
from two studies that have one treatment in common (e.g. A vs B, B vs C).17 This type of
11
analysis can only be applied to connected networks of randomised controlled trials (RCTs),19
12
but preserves the within-trial randomised comparison of each study19 and allows information
13
on treatment strategies to be “borrowed” from other studies within the network, thereby
14
increasing the total sample size.20,21 Network meta-analysis conducted using Bayesian
15
methods22-24 also allows the treatment strategies to be ranked in terms of clinical
16
effectiveness with an estimate of the probability that each strategy is ‘best’.25
TE D
M AN U
SC
7
17
Our primary aims were to simultaneously compare the clinical effectiveness of different
19
treatment strategies for sciatica using network meta-analyses, in order to identify the best
20
treatment and to provide estimates for all possible pairwise comparisons, based on both
21
direct and indirect evidence. Our secondary aims were to demonstrate the feasibility of using
22
network meta-analyses as a rational basis for clinical decision making when a number of
23
treatment options are available and where a series of conventional systematic reviews have
24
failed to help with real-world treatment decisions. The analyses presented in this paper
25
represent a refinement of initial network meta-analyses conducted as part of a broader
26
Health Technology Assessment (HTA) evaluating the clinical and cost effectiveness of
27
treatments for sciatica. A full account of the study methods and literature search are
28
presented in the HTA monograph (which also includes the protocol).16
AC C
EP
18
3
ACCEPTED MANUSCRIPT
1 METHODS
3
Search strategy
4
Included studies were identified via an extensive literature search described in full, including
5
the search strategy, in the HTA monograph.16 The search incorporated 28 electronic
6
databases and trial registries including MEDLINE, EMBASE, and AMED. Databases were
7
searched from inception until December 2009 without language restriction. The reference
8
lists of previous systematic reviews and included studies were also scanned for further
9
references.
SC
RI PT
2
M AN U
10 Study selection and data extraction
12
This review included any comparative study (experimental or observational) with adults who
13
had sciatica diagnosed clinically, or where clinical imaging confirmed lumbar disc prolapse
14
consistent with the clinical findings. The essential clinical criterion was radicular leg pain
15
worse than back pain.16 Studies of sciatica caused by conditions other than a prolapsed
16
intervertebral disc were included if it was documented that radicular leg pain was worse than
17
back pain. If imaging was used, it had to demonstrate evidence of nerve root compromise.
18
Studies that included participants with non-specific low back pain were only included if the
19
findings for patients with sciatica were reported separately. Any type of intervention to treat
20
sciatica was considered. These were categorised, for the purpose of the present analyses,
21
into one of 21 categories (See Table 1). Interventions that included a combination of more
22
than one treatment strategy (or mixed treatments) were excluded from the network meta-
23
analyses due to uncertainty regarding the extent of interaction between the combined
24
interventions. The same applied to post-surgical interventions due to surgery being included
25
as a separate treatment category. Studies comparing interventions that were grouped under
26
the same treatment strategy were also excluded. Three further studies evaluating
27
experimental interventions for sciatica (common peroneal nerve block,26 protolytic enzyme,28
28
and colchicine30) were excluded from the analyses as these interventions did not fit the
AC C
EP
TE D
11
4
ACCEPTED MANUSCRIPT
1
treatment categorisation. For the present network meta-analyses we concentrated on overall
2
response and pain intensity rather than back specific function, so three studies which only
3
reported outcome data for back specific function were excluded.32-34
5
RI PT
4 [Table 1: Treatment categorisation]
6
Two reviewers screened studies for inclusion independently. Data were extracted by one
8
reviewer and checked by a second using the original paper, whilst quality assessment was
9
done by two reviewers independently. Any disagreements were resolved by discussion. The
SC
7
quality of both trials and observational studies was assessed using the same checklist,
11
which was based on one used by the Back Review Group of the Cochrane Collaboration for
12
RCTs35 and another recommended by the Guidelines for Systematic Reviews in Health
13
Promotion and Public Health Taskforce36 (developed by the Effective Public Health Practice
14
Project, Canada37). The criteria covered external validity, selection bias and confounding,
15
detection bias, performance bias, and attrition bias. Studies were coded as strong, moderate
16
or weak for each domain, estimating the risk of bias.
17
TE D
M AN U
10
Outcome measures
19
Overall response or global effect was analysed as a binary outcome (treatment success vs
20
failure) and synthesised using odds ratios (ORs). Where studies reported overall response in
21
terms of both overall improvement and improvement in leg pain, the data on overall
22
improvement were used. For studies that reported both physician and patient perceived
23
global effect, the data for patients’ perceived effect were used.
AC C
24
EP
18
25
Pain intensity (on a scale of 0-100) was analysed as a continuous outcome measure using
26
weighted mean difference (WMD). We only included pain assessment from one location
27
from each study using the preference hierarchy of leg pain then overall pain. Where feasible,
28
missing data were estimated from the published data, using standard methods, such as
5
ACCEPTED MANUSCRIPT
standard deviations (SDs) derived from standard errors (SEs).38 Where mean values were
2
unavailable but the medians were reported, these were used instead. If SDs for baseline
3
values were available these were substituted for missing SDs. For studies that did not report
4
sufficient data to derive the SDs, they were imputed using the weighted mean,39 which was
5
calculated separately for each intervention category.
RI PT
1
6 Statistical analysis
8
The network meta-analyses were based on a single time point, using the findings from
9
individual studies closest to six months follow-up. Sensitivity analyses were conducted to
10
assess the impact of excluding non-randomised studies (observational studies and non-
11
RCTs).
M AN U
SC
7
12
The network meta-analyses were conducted using a hierarchical random-effects model18
14
within the Bayesian framework. Bayesian methods are based on the idea that unknown
15
quantities, such as population means or proportions, have probability distributions.23 You
16
start with a distribution that is based on prior knowledge or subjective belief about the
17
population and then update this using data from your included studies. However, using non-
18
informative priors (such as, a normal distribution with a large variance) means that the
19
results are based predominantly on the data from the included studies, and as such will
20
mirror those obtained using frequentist or classical meta-analysis methods. Bayesian
21
methods are implemented using model-based simulations, which means that they can be
22
used to perform complex analyses that incorporate multiple data sources and allow for
23
various parameter uncertainties within a single coherent model, which is why we chose to
24
use these methods.
AC C
EP
TE D
13
25 26
Our network meta-analyses were conducted using WinBUGS1.4.3 software,40 which uses
27
Markov chain Monte Carlo (MCMC) simulation methods to run thousands of simulated
28
iterations based on the data and description of the proposed distributions for relevant
6
ACCEPTED MANUSCRIPT
parameters. The iterative simulations area generally started at multiple points in order to
2
ensure the samples are drawn from the whole sampling frame. The first 50,000 iterations (or
3
burn-in) were discarded, and the results are based on a further sample of at least 100,000
4
simulations, ensuring that the multiple simulation strings have converged and distributions
5
were informed by later simulations. Numerical methods such as the Brooks-Gelman-Rubin
6
statistic40 and the inspection of the auto-correlation and history plots, which are routine
7
assessments made when using MCMC methods, were used to check that convergence had
8
occurred. The model fit was checked by the global goodness of fit statistic, residual
9
deviance. If the model is an adequate fit, it is expected that the residual deviance should be
SC
RI PT
1
roughly equal to the number of data points.19 Non-informative priors were used for normal
11
distributions for means, and uniform distributions for standard deviations. The treatment
12
strategy ‘inactive control’ was used as the reference treatment. This included interventions
13
that represent the non (active) treatment of sciatica, such as no treatment, sham treatment,
14
or placebo (two studies used active placebo). The WinBUGS codes (or models) that we
15
used are presented in Supplementary Material (Web Appendix A). The robustness of the
16
network meta-analyses were also evaluated by comparing the findings (where head to head
17
studies were available) with those of standard ‘direct’ pairwise meta-analyses16 conducted
18
using a random-effects model41 based on frequentist methods22-24 in Stata 10.
TE D
EP
19
M AN U
10
The assumptions of a random-effects network meta-analysis are that (1) the treatment
21
effects are additive (i.e. the relative effect of treatment A vs C can be estimated from the
22
effect of A vs B and B vs C);19,42,43 (2) study-specific treatment effects are drawn from a
23
common distribution (exchangeable);19,44 and (3) this common distribution or heterogeneity is
24
constant between the different comparisons.19,44 We evaluated heterogeneity between
25
studies, defined as the variability of the results across studies within each treatment
26
comparison over and above chance,45 by examining the findings of standard pairwise meta-
27
analyses using visual inspection of the forest plots, as well as Chi2 statistic to test for the I2
28
statistic to quantify statistical heterogeneity.46,47
AC C
20
7
ACCEPTED MANUSCRIPT
1 2 RESULTS
4
Included studies
5
As seen in Figure 1, 122 studies were included in the revised network meta-analyses48-168
6
(one publication included two studies122), 86 were RCTs51,52,54,55,57,59-65,67,69,70,72-74,76-79,81-86,88,90-
7
93,96,98-100,102-104,106,108-110,113,114,117-127,129-131,133,135-144,146-148,152-158,161,162,164,167
8
RCTs.49,94,107,112 The network meta-analysis of global effect included 95 studies (68 RCTs/Q-
9
RCTs) and pain intensity 53 studies (46 RCTs/Q-RCTs). A description of the interventions,
RI PT
3
SC
and four Q-
populations, study design, and outcome data for the pairwise studies are presented in
11
Supplementary Material (Web Appendix B).
M AN U
10
12
Eleven (9%) studies had a strong overall quality rating61,82,100,102,119,131,135,143,155,158,164 and
14
eight (7%) had a strong overall external validity rating;100,103,119,124,131,143,156,158 five (4%) of
15
which had a strong rating for both.100,119,131,143,158 Only 26 (21%) studies used both adequate
16
randomisation and adequate or partially adequate (using sealed envelopes, n=16) allocation
17
concealment.
TE D
13
18
The proportion of studies that limited inclusion to patients with acute sciatica (duration of
20
symptoms 10),
18
but these were not statistically significant: chemonucleolysis (E), traction (H), exercise
19
therapy (K), passive physical therapy (such as ultrasound and transcutaneous electrical
20
nerve stimulation) (L), bed rest (N), opioid medication (O), percutaneous discectomy (Q),
21
intradiscal injections (S), and radio frequency treatment (U), all of which were associated
22
with very wide confidence intervals. This reflects the limited evidence available for biological
23
agents, which included a small placebo controlled RCT (n=24) that reported a large effect
24
estimate in favour of biological agents (OR 10.0; 95% CI: 0.65, 166.67. see Supplementary
25
Material Table C1).
AC C
EP
TE D
M AN U
10
26 27
The results of the sensitivity analyses excluding observational studies and non-RCTs
28
showed broad agreement with the main analyses. For global effect, the most notable
10
ACCEPTED MANUSCRIPT
discrepancies occurred with biological agents compared with chemonucleolysis,
2
conventional, and care exercise therapy. A more detailed narrative of the differences
3
between the analyses with and without the non-randomised studies is presented in the
4
Supplementary Material (Web Appendix D)
RI PT
1
5 Pain intensity
7
In terms of pain intensity, the only treatment comparisons with inactive control that were
8
statistically significant were epidural injections (D) and biological agents (M). Biological
9
agents, which had the highest probability of being best (0.33), were also found to be
SC
6
statistically significantly better at reducing pain than non-opioids (F), bed rest (N), opioids (O)
11
and radio frequency treatment (U); these findings were all associated with wide credible
12
intervals. When considering the magnitude of effect, bed rest (N), education/advice alone
13
(P), percutaneous discectomy (Q), and radiofrequencly treatment (U) tended to fare worse
14
when compared with most treatment strategies, with findings showing a non-statistically
15
significant difference of more than 25 points. Acupuncture (J), had the second highest
16
probability of being best (0.19) and resulted in reductions of pain intensity of more than 25
17
points compared with bed rest, opioids, education/advice alone, percutaneous discectomy
18
and radio frequency treatment, none of which were statistically significant and all had wide
19
credible intervals.
TE D
EP
20
M AN U
10
For pain intensity the most notable discrepancies between the network meta-analysis with
22
and without observational studies and non-RCTs only occurred with biological agents (vs
23
inactive control, conventional care, disc surgery, non-opioids, intra-operative interventions,
24
acupuncture, exercise therapy, opioids, and neuropathic painmodulators). Biological agents
25
no longer had the highest probability of being best (0.03; see Supplementary material Table
26
C4). These discrepancies are likely to be due to the small number of included studies with a
27
limited number of participants evaluating biological agents (2 RCTs n=131; 1 non-
28
randomised RCT n=72; and 1 historical cohort study n=10).
AC C
21
11
ACCEPTED MANUSCRIPT
1 2 Between study heterogeneity, model fit and comparison with standard pairwise meta-
4
analyses
5
Based on the Gelman-Rubin statistic, convergence occurred at around 6-8000 iterations for
6
both outcome measures (global effect, pain intensity). The auto-correlation and history plots
7
also showed good convergence. The goodness of fit of the models to the data, measured by
8
the residual deviance, was found to be good for all three outcomes (Supplementary Material,
9
Web Appendix E).
SC
RI PT
3
M AN U
10
The results of the evaluation of between-study heterogeneity showed a moderate to high
12
level16 of statistical heterogeneity for many of the pairwise comparisons, as well as across all
13
studies as a whole. The heterogeneity was greater for the analysis of pain intensity than
14
global effect, with an I2 statistic of less than 75% (i.e. moderate or less) for all but one
15
pairwise comparison (epidural injections vs conventional care). The observed values for I2
16
are presented in Figure 1. Heterogeneity did not improve when non-randomised studies
17
were removed.
TE D
11
18
The comparison of the results from the network meta-analyses with that of the conventional
20
pairwise meta-analyses showed broad agreement with slightly more discrepancies for the
21
analyses of pain intensity. These discrepancies were greatest for comparisons that had very
22
little direct evidence, such as biological agents.
24
AC C
23
EP
19
25
DISCUSSION
26
This is the first systematic review that has included all treatment strategies for sciatica in the
27
same analysis using a network meta-analysis method that includes indirect comparisons.
28
The advantages of such analyses are that they can simultaneously compare more than two
12
ACCEPTED MANUSCRIPT
treatments in the same coherent analysis; provide relative effect estimates for all treatment
2
comparisons, even those that have not been directly compared in head to head trials; enable
3
the estimation of the probability that each treatment is best; and reduce the uncertainty in the
4
treatment effect estimates.
RI PT
1
5 Summary of results
7
In terms of overall response or global effect, there was a statistically significant improvement
8
following disc surgery, epidural injections, non-opioid medication, intra-operative
9
interventions, manipulation, and acupuncture when compared with inactive control or
SC
6
conventional care. Epidural injections, disc surgery, and intra-operative interventions
11
were also statistically significantly better than percutaneous discectomy, chemonucleolysis,
12
intradiscal injections, and radiofrequencly treatment; with epidural injections, and intra-
13
operative interventions also statistically significantly better than both traction, and exercise
14
therapy. While biological agents and acupuncture had the highest probability of being best
15
and had the largest effect estimates when compared with inactive control, these findings
16
were associated with very wide credible intervals, reflecting the lack of information on these
17
effect estimates.
TE D
M AN U
10
EP
18
In terms of pain intensity, there was a statistically significant reduction in pain following
20
epidural injections and biological agents compared with inactive control, but there was no
21
significant difference between disc surgery and inactive control. Biological agents had the
22
highest probability of being best, and were also statistically significantly better than non-
23
opioid medication, opioid medication, bed rest, and radio frequency treatment. However,
24
when the analysis was restricted to RCTs, biological agents no longer had the highest
25
probability of being best and were not found to be statistically better than any other
26
treatments. When considering the magnitude of effect, bed rest, education/advice alone,
27
percutaneous discectomy, and radiofrequencly treatment were considerably inferior when
AC C
19
13
ACCEPTED MANUSCRIPT
1
compared with most treatment strategies, but these findings were not statistically significant
2
and were associated with wide credible intervals.
3 Overall, the results of the sensitivity analyses excluding non-randomised studies showed
5
broad agreement with the main analyses, with the findings generally becoming non-
6
statistically significant due to broader credible intervals for the analyses restricted to RCTs
7
and Q-RCTs. The most notable discrepancies occurred with treatment strategies that were
8
associated with a small number of included studies such as those reporting treatment with
9
biological agents.
SC
RI PT
4
M AN U
10 Findings of previous reviews
12
Previous reviews of non-surgical treatments have either found no evidence of
13
effectiveness,9,10 conflicting evidence,11,12 or have reached different conclusions concerning
14
the effectiveness of epidural steroid injections.9,11,13,14,169,170 A Cochrane systematic review of
15
surgical interventions did not combine the results of four RCTs comparing discectomy with
16
non-surgical treatment due to heterogeneity, and concluded that the results showed a
17
temporary benefit of disc surgery at one year follow-up.15 In that review the effectiveness of
18
discectomy was justified by using informal indirect comparison of chemonucleolysis with
19
placebo, and chemonucleolysis with disc surgery; chemonucleolysis was more effective than
20
placebo and discectomy more effective than chemonucleolysis, therefore disc surgery was
21
superior to placebo. Using our network meta-analyses, it was possible to make a more
22
robust statement on disc surgery compared with placebo: disc surgery was statistically
23
significantly better than placebo in terms of global effect but not for pain intensity.
24
AC C
EP
TE D
11
25
Strengths and weaknesses
26
One of the main strengths of our network meta-analyses is the wide range of treatment
27
strategies used to treat sciatica that were not only considered in the same review, but
28
compared simultaneously in the same analysis. Another strength is that they were based on
14
ACCEPTED MANUSCRIPT
a systematic and comprehensive search of the literature up (until December 2009) that
2
covered any therapeutic intervention for sciatica. Although we acknowledge that these
3
searches are not current, and as such, more recent relevant data is likely to have been
4
excluded.
RI PT
1
5
The RCT is widely regarded as the design of choice when assessing the effectiveness of
7
health care interventions171 and we acknowledge the controversy over the inclusion of non-
8
randomised evidence. Non-randomised studies were included in the search because some
9
treatment approaches may not have been evaluated by RCTs, and also to increase the
SC
6
precision of the findings for interventions evaluated by a limited number of studies.
11
Observational studies can have better external validity than RCTs172,173 and provide more
12
generalisable findings. However, observational studies are likely to be affected by selection
13
bias and confounding, and may therefore yield estimates of association that deviate from the
14
true underlying relationship beyond the play of chance.174 As it happens, most of the RCTs
15
did not report the method of generating the randomisation sequence or allocation
16
concealment, which means that selection bias or confounding might still be present.
17
Excluding the non-randomised studies in a sensitivity analysis did not affect the structure of
18
the network and the overall findings of both series of network meta-analyses were similar,
19
although less precise for the analyses of RCTs.
TE D
EP
20
M AN U
10
Network meta-analysis methods enabled us to go beyond the pairwise comparisons reported
22
in previous systematic reviews. They allowed us to simultaneously compare all the available
23
treatment strategies for sciatica and provided estimates of relative treatment effects for all
24
conceivable comparisons, even those where there was no direct evidence available.
25
However, the small number of relevant studies for some comparisons, statistical
26
heterogeneity (within pairwise comparisons), and potential inconsistency (between pairwise
27
comparisons) within the networks means that the encouraging results for interventions such
28
as biological agents should be interpreted with caution.
AC C
21
15
ACCEPTED MANUSCRIPT
1 In order to answer the question of which is the optimum treatment for sciatica and provide
3
generalisable findings, we were interested in the average treatment effect of each treatment
4
approach (to represent the diversity used in clinical practice). We therefore pooled clinically
5
heterogeneous studies. We used a random-effects model to pool the data, which is based
6
on the assumption that different studies assessed different, yet related, treatment effects.
7
However, included studies also varied in study design and risk of bias (methodological
8
diversity). There was considerable (I2 ≥75%)46 statistically significant between-study
9
heterogeneity present for a number of comparisons within the pairwise meta-analyses,
SC
RI PT
2
especially in the analyses of pain intensity, and it was not possible to ascertain how much
11
was due to clinical or methodological diversity. This needs to be taken into consideration in
12
future work.
M AN U
10
13
The network meta-analyses relied on the key assumption that the relative treatment effect of
15
one treatment versus another is the same across the entire set of studies.18,44 The use of
16
random-effects models meant that it was assumed that the common distribution of effects
17
was the same across all sets of studies. A further assumption made in the analyses was that
18
the relative efficacy of different treatments is the same at different stages in the care
19
pathway. Pragmatically, sciatica is often treated with a stepped care approach starting with
20
conservative treatments, such as non-opioid medication, progressing if necessary to more
21
invasive treatments such as epidural injections or surgery. This means that the population of
22
patients treated with conservative treatments was likely to differ from those treated with
23
invasive treatments, resulting in confounding and inconsistency within the network. Although
24
descriptive characteristics were generally poorly reported by included studies, there was a
25
trend for studies evaluating invasive treatments to report a history of previous treatments
26
and include patients with a diagnosis confirmed by imaging, and for studies of conservative
27
treatments to limit inclusion to patients with acute sciatica. Due to the breadth of the review
28
and the novel and speculative use of network meta-analysis methods, we have not yet
AC C
EP
TE D
14
16
ACCEPTED MANUSCRIPT
1
incorporated stepped care approaches in the network meta-analyses. The optimum
2
sequence of treatment modalities and what sequence is best for which patients is therefore
3
not yet known and awaits further analysis.
RI PT
4 The network meta-analyses were based on a single time point, outcome data closest to six
6
months, which may be considered as a limitation of the analyses. The HTA monograph16
7
included an assessment of each treatment strategy at short (≤ 6 weeks), medium (> 6weeks
8
to ≤ 6 months) and long (> 6 months) term follow-up, but this evaluation was based on
9
multiple pairwise analyses, with each analysis needing to be interpreted independently.
10
Further research is needed to incorporate multiple time points within the network meta-
11
analyses in order to incorporate data at different follow-up periods.
M AN U
SC
5
12
For the pain intensity outcome, where the SDs were missing (and could not be estimated
14
from the published data) these were imputed using the weighted mean SD39,175 for each
15
treatment strategy (11 studies). This is based on the assumption that the variance is similar
16
between studies and the data are not skewed.176 We also used medians to represent the
17
mean for two studies. We considered that it was better to use these methods in order to
18
incorporate more of the evidence base, as ignoring the findings of these studies may induce
19
bias in the summary effect estimate.175 Furukawa, et al.39 have previously shown that it is
20
safe to borrow SDs from other studies.
EP
AC C
21
TE D
13
22
There were insufficient studies to explore the presence of publication or reporting bias for
23
most treatment comparisons. However, a funnel plot of studies comparing surgery and
24
chemonucleolysis showed no evidence of publication bias.16 The benefit (or effectiveness) of
25
different treatment strategies for sciatica should be considered along with potential harms.
26
Although the present paper does not report adverse effects, they are reported elsewhere.16
27 28
CONCLUSIONS
17
ACCEPTED MANUSCRIPT
The use of network meta-analyses has enabled us to provide new information on the relative
2
effectiveness of treatments for sciatica. This can help clinicians and patients in shared
3
decision making, as well as providing data for healthcare policy development. The findings
4
provide support for the effectiveness of some common therapies for sciatica such as non-
5
opioid medication, epidural injections and disc surgery. They also suggest that less
6
frequently used treatments such as manipulation and acupuncture, and experimental
7
treatments such as cytokine modulating biological agents, may be considered. The findings
8
of this review do not support the effectiveness of opioid medication, either for pain intensity
9
or global effect. Furthermore, there is no support for the effectiveness of numerous other
SC
RI PT
1
interventions such as bed rest, exercise therapy, percutaneous discectomy or traction. The
11
lack of support for education/advice should not be taken to imply that patients should not be
12
given information or advice; rather it is not an effective treatment if delivered alone.
M AN U
10
13
Further research is needed to confirm or refute these findings where we found limited
15
evidence, and to explore the impact of heterogeneity and the range of clinical questions
16
most suited for the use of network meta-analyses. There is also scope to develop more
17
sophisticated methods, such as building on the confidence profile method,173 bias-adjusted
18
results,177 or Bayesian statistics,_ENREF_160172 to incorporate information relating to
19
differences in study design or internal and external validity in the network meta-analyses, as
20
well as data on multiple follow-up periods. The issue of how best to estimate the
21
effectiveness of treatment approaches according to their order within a sequential treatment
22
pathway remains an important challenge.
24 25
EP
AC C
23
TE D
14
REFERENCES
18
ACCEPTED MANUSCRIPT
1
1.
Deyo RA, Rainville J, Kent DL. What can the history and physical examination tell us
2
about low back pain? Journal of the American Medical Association. 1992;268(6):760-
3
765. 2.
Bush K, Cowan N, Katz DE, et al. The natural history of sciatica associated with disc
RI PT
4 5
pathology; a prospective study with clinical and independent radiological follow-up.
6
Spine 1992;18:1433-1438. 3.
8
Journal of Clinical Epidemiology 2004;57:174-179. 4.
10 11
Koes BW, van Tulder MW, Peul WC. Clinical Review: Diagnosis and treatment of sciatica BMJ. 2007;334:1313-1317.
5.
SC
9
Tubach F, Beaute J, Leclerc A. Natural history and prognostic indicators of sciatica.
M AN U
7
Weber H, Holme I, Amlie E. The natural course of acute sciatica with nerve root
12
symptoms in a double-blind placebo-controlled trial evaluating the effect of piroxicam.
13
Spine 1993;18:1433-1438. 6.
15 7.
17 18
8.
9.
Vroomen PC, de Krom MC, Slofstra PD, Knottnerus JA. Conservative treatment of sciatica: a systematic review. J Spinal Disord. Dec 2000;13(6):463-469.
10.
23 24
Konstantinou K, Dunn KM. Sciatica: Review of epidemiological studies and prevalence estimates. Spine 2008;33:2464-2472.
21 22
Samanta A, Beardsley J. Evidence based case report: Sciatica: which intervention? . BMJ 1999;319:302-303.
19 20
TE D
canal. . N Engl J Med 1934;211:210-215.
EP
16
Mixer WJ, Barr JS. Rupture of the intervertebral disc with involvement of the spinal
AC C
14
Hagen KB, Hilde G, Jamtvedt G, Winnem M. Best rest for acute low back pain and sciatica. The Cochrane Database of Systematic Reviews 2004.
11.
Luijsterburg PAJ, Verhagen AP, Ostelo RWJG, van Os TAG, Peul WC, Koes BW.
25
Effectiveness of conservative treatments for the lumbosacral radicular syndrome: a
26
systematic review. European Spine Journal. Jul 2007;16(7):881-899.
27 28
12.
Clarke JA, van Tulder MW, Blomberg SEI, et al. Traction for low-back pain with or without sciatica. Cochrane Database of Systematic Reviews. 2007(2).
19
ACCEPTED MANUSCRIPT
1
13.
Boswell MV, Hansen HC, Trescot AM, Hirsch JA. Epidural steroids in the
2
management of chronic spinal pain and radiculopathy. Pain Physician. Jul
3
2003;6(3):319-334.
5
spinal pain: a systematic review. Pain Physician. Jan 2007;10(1):185-212. 15.
7 8
Database of Systematic Reviews. 2007(2). 16.
9
Lewis R, Williams N, Matar HE, et al. The effectiveness and cost effectiveness of
management strategies for sciatica: systematic review and economic model. Health
10 11
Gibson JNA, Waddell G. Surgical interventions for lumbar disc prolapse. Cochrane
SC
6
Abdi S, Datta S, Trescot AM, et al. Epidural steroids in the management of chronic
RI PT
14.
Technology Assessment. 2011;15(39). 17.
M AN U
4
Psaty BM, Lumley T, Furberg CD, et al. Health Outcomes Associated With Various
12
Antihypertensive Therapies Used as First-Line Agents: A Network Meta-analysis
13
Jama. 2003;289(19):2534-2544 18.
15 16
Lu G, Ades AE. Combination of direct and indirect evidence in mixed treatment comparisons. Statistics in Medicine. 2004;23(15449338):3105-3124.
19.
TE D
14
Cooper NJ, Sutton AJ, Lu G, Khunti K. Mixed Comparison of Stroke Prevention
17
Treatments in Individuals With Nonrheumatic Atrial Fibrillation Archives of Internal
18
Medicine. 2006;166:1269-1275. 20.
Anothaisintawee T, Attia J, Nickel JC, et al. Management of Chronic Prostatitis/
EP
19
Chronic Pelvic Pain Syndrome: A Systematic Review and Network Meta-analysis
21
JAMA. 2011;305 (1):78-86.
22
21.
23 24
27
Higgins JPT, Whitehead A. Borrowing strength from external trials in a meta-analysis. Statistics in Medicine 1996;15:2733-2749.
22.
25 26
AC C
20
Sutton AJ, Abrams KR. Bayesian methods in meta-analyis and evidence synthesis. Statistical Methods in Medical Research. 2001;10:277-303.
23.
Bland MJ, Altman DG. Statistics notes: Bayesians and frequentists. BMJ. October 24 1998;317(166):1151–1160.
20
ACCEPTED MANUSCRIPT
1
24.
2 3
Ashby D, Smith AFM. Evidence-based medicine as Bayesian decision-making. Statistics in Medicine. 2000;19:3291-3305.
25.
Salanti G, Ades AE, Ioannidis JPA. Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial.
5
Journal of Clinical Epidemiology. 2011;64:163-171.
6
26.
RI PT
4
Tajiri K, Takahashi K, Ikeda K, Tomita K. Common peroneal nerve block for sciatica. Clinical Orthopaedics and Related Research. 1998(347):203-207.
8
http://www.mrw.interscience.wiley.com/cochrane/clcentral/articles/952/CN-
9
00148952/frame.html 27.
11 12
Hedges LV, Tipton E, Johnson ML. Robust variance estimation in meta-regression
M AN U
10
SC
7
with dependent effect size estimates. Research Synthesis Methods. 2010;1:39-65. 28.
13
Gibson T, Dilke TF, Grahame R. Chymoral in the treatment of lumbar disc prolapse. Rheumatology & Rehabilitation. Aug 1975;14(3):186-190.
29.
Bland M. Meta-analysis: publication bias. 2006.
15
30.
El-Zahaar MS. The Egyptian experience in the use of colchicine in lumbar disc
TE D
14
16
disease. Journal of Neurological and Orthopaedic Medicine and Surgery.
17
1995;16(3):147-152.
19 20
165. 32.
21
radiculopathy. Spine. Jun 1 2007;32(13):1414-1422.
33.
24
Lavyne MH, Bilsky MH. Epidural steroids, postoperative morbidity, and recovery in patients undergoing microsurgical lumbar discectomy. Journal of Neurosurgery. Jul
25 26
Thomas KC, Fisher CG, Boyd M, Bishop P, Wing P, Dvorak MF. Outcome evaluation of surgical and nonsurgical management of lumbar disc protrusion causing
22 23
Gelman A. Bayesian Statistics Then and Now. Statistical Science. 2010;25(2):162-
EP
31.
AC C
18
1992;77(1):90-95. 34.
Ejeskar A, Nachemson A, Herberts P, et al. Surgery versus chemonucleolysis for
27
herniated lumbar discs. A prospective study with random assignment. Clin Orthop.
28
Apr 1983(174):236-242.
21
ACCEPTED MANUSCRIPT
1
35.
van Tulder MW, Furlan A, Bombardier C, Bouter LM, Editorial Board of the Cochrane Collaboration Back Review Group. Updated method guidelines for systematic
3
reviews in the Cochrane collaboration back review group. Spine. 2003;28(12):1290-
4
1299.
5
36.
RI PT
2
Jackson N, Waters E, For the Guidelines for Systematic Reviews in Health
Promotion and Public health Taskforce. Criteria for the systematic review of health
7
promotion and public health interventions. Health Promotion International.
8
2005;20(4):367-374. 37.
10 11
Effective Public Health Practice Project. Quality Assessment tool for Quantitative Studies http://www.ephpp.ca/tools.html. 2007 (Accessed 31 July, 2007).
38.
M AN U
9
SC
6
Higgins JPT, Deeks JJ. Chapter 7: Selecting studies and collecting data. In: Higgins
12
JPT, Green S, eds. Cochrane handbook for systematic reviews of interventions.
13
Chichester (UK): John Wiley & Sons Ltd; 2008.
14
39.
Furukawa TA, Barbui C, Cipriani A, Brambilla P, Watanabe N. Imputing missing standard deviations in meta-analyses can provide accurate results. J Clin Epidemiol.
16
Jan 2006;59(1):7-10. 40.
18
manual. Cambridge: MRC Biostatistics Unit; 2003. 41.
20 21
1986;7:177-188. 42.
22
Song F, Altman DG, Glenny A-M, Deeks JJ. Validity of indirect comparison for estimating efficacy of competing interventions: empirical evidence from published
23 24
DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled clinical trials
EP
19
Spiegelhalter DJ, Thomas A, Best NG, Lunn DJ, eds. WinBUGS Version v1.4. User
AC C
17
TE D
15
meta-analyses. BMJ. 2003;326:472-476.
43.
Bucher HC, Guyatt GH, Griffith LE, Walter SD. The Results of Direct and Indirect
25
Treatment Comparisons in Meta-Analysis of Randomized Controlled Trials. J Clin
26
Epidemiol. 1997;50(6):683-691.
22
ACCEPTED MANUSCRIPT
1
44.
Caldwell DM, Ades AE, Higgins JPT. Simultaneous comparison of multiple
2
treatments: combining direct and indirect evidence. BMJ. 2005;331(16223826):897-
3
900. 45.
Stettler C, Allemann S, Wandel S, et al. Drug eluting and bare metal stents in people
RI PT
4 5
with and without diabetes: collaborative network meta-analysis BMJ.
6
2008;337:a1331.
7
46.
Deeks JJ, Higgins JPT, Altman DG. Chapter 9: Analysing data and undertaking
meta-analyses. In: Higgins JPT, Green S, eds. Cochrane Handbook for Systematic
9
Reviews of Interventions. Chichester (UK): John Wiley & Sons; 208. 47.
11 12
Higgins JPT, Thompson SG, Deeks JJ, G. AD. Measuring inconsistency in meta-
M AN U
10
SC
8
analyses. BMJ. 2003;327(7414):557-560 48.
Alexander AH, Burkus JK, Mitchell JB, Ayers WV. Chymopapain chemonucleolysis
13
versus surgical discectomy in a military population. Clin Orthop Relat Res.
14
1989(244):158-165. 49.
Aminmansour B, Khalili HA, Ahmadi J, Nourian M. Effect of high-dose intravenous
TE D
15 16
dexamethasone on postlumbar discectomy pain. Spine. Oct 1 2006;31(21):2415-
17
2417.
18
50.
Atlas SJ, Deyo RA, Keller RB, et al. The Maine Lumbar Spine Study .2. 1-year outcomes of surgical and nonsurgical management of sciatica. Spine. Aug
20
1996;21(15):1777-1786. 51.
22
Efficacy of epidural perineural injections with autologous conditioned serum for
23
lumbar radicular compression: an investigator-initiated, prospective, double-blind,
24 25
Becker C, Heidersdorf S, Drewlo S, de Rodriguez SZ, Kramer J, Willburger RE.
AC C
21
EP
19
reference-controlled study. Spine. Aug 1 2007;32(17):1803-1808.
52.
Bernsmann K, Kramer J, Ziozios I, Wehmeier J, Wiese M. Lumbar micro disc surgery
26
with and without autologous fat graft - A prospective randomized trial evaluated with
27
reference to clinical and social factors. Arch. Orthop. Trauma Surg. Sep
28
2001;121(8):476-480.
23
ACCEPTED MANUSCRIPT
1
53.
Bonafe A, Tremoulet M, Sabatier J, et al. Foraminal and Lateroforaminal Herniated
2
Disks - Midterm Results of Percutaneous Treatment with Chemonucleolysis or
3
Nucleotomy. Neurochirurgie. 1993;39(2):110-115. 54.
Bontoux D, Alcalay M, Debiais F, et al. Treatment of Lumbar-Disk Herniation with
RI PT
4 5
Intradiscal Injection of Chymopapain or Triamcinolone Hexacetonid - a Comparative-
6
Study of 80 Cases. Rev Rhum Engl Ed. Apr 1990;57(4):327-331.
7
55.
Bourgeois P, Benoist M, Palazzo E, et al. Multicenter, randomized and double-blind study of triamcinolone hexacetonide versus chymopapain in the treatment of lumbar
9
sciatica. Preliminary results at six months. Rev Rhum Mal Osteo Articulaires.
10 56.
12 13 14 15
Brown MD, Tompkins JS. Pain response post-chemonucleolysis or disc excision. Spine. Mar 1989;14(3):321-326.
57.
M AN U
1988(10):767-769.
Buchner M, Zeifang F, Brocai DR, Schiltenwolf M. Epidural corticosteroid injection in the conservative management of sciatica. Clin Orthop. Jun 2000(375):149-156.
58.
Buric J. Ozone chemyonucleolysis vs microdiscectomy. Prospective controlled study
TE D
11
SC
8
16
with 18 months follow-up. Rivista Italiana di Ossigeno-Ozonoterapia. Apr
17
2005;4(1):49-54.
18
59.
Burton AK, Tillotson KM, Cleary J. Single-blind randomised controlled trial of chemonucleolysis and manipulation in the treatment of symptomatic lumbar disc
20
herniation. European Spine Journal. Jun 2000;9(3):202-207. 60.
22
plus procaine for the management of intractable sciatica. Spine. May 1991;16(5):572-
23 24
Bush K, Hillier S. A controlled study of caudal epidural injections of triamcinolone
AC C
21
EP
19
575.
61.
Carette S, Leclaire R, Marcoux S, et al. Epidural corticosteroid injections for sciatica
25
due to herniated nucleus pulposus. New England Journal of Medicine. Jun
26
1997;336(23):1634-1640.
27 28
62.
Cengiz SL, Baysefer A. Efficacy of Adcon-L gel or Healon-GV in epidural fibrosis after lumbar microdiscectomy. Neurosciences. Apr 2007;12(2):109-113.
24
ACCEPTED MANUSCRIPT
1
63.
Chatterjee S, Foy PM, Findlay GF. Report of a controlled clinical trial comparing
2
automated percutaneous lumbar discectomy and microdiscectomy in the treatment of
3
contained lumbar disc herniation. Spine. Mar 15 1995;20(6):734-738. 64.
Chen M-r, Wang P, Cheng G, Guo X, Wei G-w, Cheng X-h. The warming
RI PT
4 5
acupuncture for treatment of sciatica in 30 cases. J Tradit Chin Med. 2009;29(1):50-
6
53.
7
65.
Cohen S, Bogduk N, Dragovich A, et al. Randomized, double-blind, placebo-
controlled, dose-response, and preclinical safety study of transforaminal epidural
9
etanercept for the treatment of sciatica. Anesthesiology. 2009;110(5):1116-1126. 66.
11 12
Coomes EN. Comparison between Epidural Anaesthesia and Bed Rest in Sciatica.
M AN U
10
SC
8
British Medical Journal. 1961;1(5218):20-24. 67.
Crawshaw C, Frazer AM, Merriam WF, Mulholland RC, Webb JK. A comparison of
13
surgery and chemonucleolysis in the treatment of sciatica. A prospective randomized
14
trial. Spine. Mar 1984;9(2):195-198. 68.
16 17
Dabezies EJ, Brunet M. Chemonucleolysis vs laminectomy. Orthopedics. Jan-Feb 1978;1(1):26-29.
69.
TE D
15
Dabezies EJ, Langford K, Morris J, Shields CB, Wilkinson HA. Safety and efficacy of chymopapain (Discase) in the treatment of sciatica due to a herniated nucleus
19
pulposus. Results of a randomized, double-blind study. Spine. May 1988;13(5):561-
20
565. 70.
22 23
discectomy. J Spinal Disord Tech. Aug 2002;15(4):273-276.
71.
24 25
Debi R, Halperin N, Mirovsky Y. Local application of steroids following lumbar
AC C
21
EP
18
Dei-Anang K, Weigand H, Mader U. Is the Percutaneous Discectomy an Alternative to Chemonucleolysis. Radiologe. 1990;30(2):70-74.
72.
Dilke TF, Burry HC, Grahame R. Extradural corticosteroid injection in management of
26
lumbar nerve root compression. British Medical Journal. Jun 16 1973;2(5867):635-
27
637.
25
ACCEPTED MANUSCRIPT
1
73.
Dincer U, Kiralp MZ, Cakar E, Yasar E, Dursan H. Caudal epidural injection versus
2
non-steroidal anti-inflammatory drugs in the treatment of low back pain accompanied
3
with radicular pain. Joint Bone Spine. Oct 2007;74(5):467-471. 74.
Dreiser RL, Le Parc JM, Velicitat P, Lleu PL. Oral meloxicam is effective in acute
RI PT
4 5
sciatica: two randomised, double-blind trials versus placebo or diclofenac.
6
Inflammation Research. Mar 2001;50 Suppl 1:S17-23. 75.
8 9
Dubourg G, Rozenberg S, Fautrel B, et al. A pilot study on the recovery from paresis after lumbar disc herniation. Spine. Jul 1 2002;27(13):1426-1431; discussion 1431.
76.
SC
7
Duplan B, Cabanel G, Piton JL, Grauer JL, Phelip X. Acupuncture and sciatica in the acute phase. Double-blind study of 30 cases. Semaine des Hopitaux. Dec 8
11
1983;59(45):3109-3114.
12
77.
M AN U
10
Feldman J, Menkes CJ, Pallardy G, et al. Double-blind study of the treatment of disc
13
lumbosciatica by chemonucleolysis. Rev Rhum Mal Osteoartic. Mar 1986;53(3):147-
14
152. 78.
Finckh A, Zufferey P, Schurch M-A, Balague F, Waldburger M, So AKL. Short-term
TE D
15 16
efficacy of intravenous pulse glucocorticoids in acute discogenic sciatica. A
17
randomized controlled trial. Spine. Feb 15 2006;31(4):377-381.
18
79.
Gallucci M, Limbucci N, Zugaro L, et al. Sciatica: treatment with intradiscal and intraforaminal injections of steroid and oxygen-ozone versus steroid only. Radiology.
20
Mar 2007;242(3):907-913. 80.
22 23
severe sciatica: a pilot study. Ann Rheum Dis. Sep 2004;63(9):1120-1123.
81.
24
Gerszten PC, Moossy JJ, Flickinger JC, Welch WC. Low-dose radiotherapy for the inhibition of peridural fibrosis after reexploratory nerve root decompression for
25 26
Genevay S, Stingelin S, Gabay C. Efficacy of etanercept in the treatment of acute,
AC C
21
EP
19
postlaminectomy syndrome. Journal of Neurosurgery. Oct 2003;99(3 Suppl):271-277. 82.
Geurts JWM, van Wijk R, Wynne HJ, et al. Radiofrequency lesioning of dorsal root
27
ganglia for chronic lumbosacral radicular pain: a randomised, double-blind, controlled
28
trial. Lancet. Jan 2003;361(9351):21-26.
26
ACCEPTED MANUSCRIPT
1
83.
Ghoname EA, White PF, Ahmed HE, Hamza MA, Craig WF, Noe CE. Percutaneous
2
electrical nerve stimulation: an alternative to TENS in the management of sciatica.
3
Pain. Nov 1999;83(2):193-199. 84.
Glasser RS, Knego RS, Delashaw JB, Fessler RG. The perioperative use of
RI PT
4 5
corticosteroids and bupivacaine in the management of lumbar disc disease. Journal
6
of Neurosurgery. Mar 1993;78(3):383-387. 85.
8
1992;17(4):388-394. 86.
10 11
Goldie I. A clinical trial with indomethacin (indomee(R)) in low back pain and sciatica. Acta Orthop Scand Suppl. 1968;39(1):117-128.
87.
SC
9
Gogan WJ, Fraser RD. Chymopapain - a 10-Year, Double-Blind-Study. Spine. Apr
M AN U
7
Graham CE. Chemonucleolysis: a double blind study comparing chemonucleolysis
12
with intra discal hydrocortisone: in the treatment of backache and sciatica. Clin
13
Orthop. Jun 1976(117):179-192. 88.
15 16
Zeitschrift fur Rheumatologie. Nov-Dec 1975;34(11-12):444-447. 89.
17 18
Grevsten S, Johansson H. Phenylbutazone in treatment of acute lumbago-sciatica.
TE D
14
Hansson E, Hansson T. The cost-utility of lumbar disc herniation surgery. European Spine Journal. Mar 2007;16(3):329-337.
90.
Hedeboe J, Buhl M, Ramsing P. Effects of using dexamethasone and placebo in the treatment of prolapsed lumbar disc. Acta Neurologica Scandinavica. Jan
20
1982;65(1):6-10. 91.
22
sciatica by a single extradural corticosteroid injection. Br J Clin Pract.
23 24
Helliwell M, Robertson JC, Ellis RM. Outpatient treatment of low pack pain and
AC C
21
EP
19
1985;39(6):228-231.
92.
Herrmann WA, Geertsen MS. Efficacy and safety of lornoxicam compared with
25
placebo and diclofenac in acute sciatica/lumbo-sciatica: an analysis from a
26
randomised, double-blind, multicentre, parallel-group study. Int J Clin Pract.
27
2009;63(11):1613-1621.
27
ACCEPTED MANUSCRIPT
1
93.
Hofstee DJ, Gijtenbeek JMM, Hoogland PH, et al. Westeinde sciatica trial:
2
randomized controlled study of bed rest and physiotherapy for acute sciatica. Journal
3
of Neurosurgery. Jan 2002;96(1 Suppl):45-49. 94.
5 6
Holve RL, Barkan H. Oral steroids in initial treatment of acute sciatica. J Am Board
RI PT
4
Fam Med. 2008;21(5):469-474. 95.
Hoogmartens M, Van Loon L, Demedts D, Desmet L. Comparative study of the
results obtained with discectomy and chemonucleolysis. Acta Orthop Belg. Mar-Apr
8
1976;42(2):134-138.
9
96.
SC
7
Javid MJ, Nordby EJ, Ford LT, et al. Safety and efficacy of chymopapain (Chymodiactin) in herniated nucleus pulposus with sciatica. Results of a randomized,
11
double-blind study. Journal of the American Medical Association. May 13
12
1983;249(18):2489-2494.
13
97.
14
Javid MJ. Chemonucleolysis Versus Laminectomy - a Cohort Comparison of Effectiveness and Charges. Spine. Sep 1995;20(18):2016-2022.
98.
Jensen TT, Asmussen K, Berg-Hansen EM, et al. First-time operation for lumbar disc
TE D
15
M AN U
10
16
herniation with or without free fat transplantation. Prospective triple-blind randomized
17
study with reference to clinical factors and enhanced computed tomographic scan 1
18
year after operation. Spine. May 1 1996;21(9):1072-1076. 99.
Jirarattanaphochai K, Jung S, Thienthong S, Krisanaprakornkit W, Sumananont C.
EP
19
Peridural methylprednisolone and wound infiltration with bupivacaine for
21
postoperative pain control after posterior lumbar spine surgery - A randomized
22 23
double-blinded placebo-controlled trial. Spine. 2007;32(6):609-616.
100.
24 25
AC C
20
Karppinen J, Malmivaara A, Kurunlahti M, et al. Periradicular infiltration for sciatica A randomized controlled trial. Spine. May 2001;26(9):1059-1067.
101.
Karppinen J, Korhonen T, Malmivaara A, et al. Tumor necrosis factor-alpha
26
monoclonal antibody, infliximab, used to manage severe sciatica. Spine. Apr 15
27
2003;28(8):750-753; discussion 753-754.
28
ACCEPTED MANUSCRIPT
102.
2 3
chronic lumbar radicular pain. The Journal of Pain. 2005;6(12):829-836. 103.
4 5
Khoromi S, Patsalides A, Parada S, Salehi V, Meegan JM, Max MB. Topiramate in
Khoromi S, Cui L, Nackers L, Max MB. Morphine, nortriptyline and their combination vs. placebo in patients with chronic lumbar root pain. Pain. Jul 2007;130(1-2):66-75.
104.
RI PT
1
Kim KD, Wang JC, Robertson DP, et al. Reduction of radiculopathy and pain with
6
Oxiplex/SP gel after laminectomy, laminotomy, and discectomy: a pilot clinical study.
7
Spine. May 15 2003;28(10):1080-1087; discussion 1087-1088.
9
endoscopic diskectomy in 295 patients: comparison with results of microscopic
10
diskectomy. Surg Neurol. Dec 2007;68(6):623-631. 106.
12 13
Klenerman L, Greenwood R, Davenport HT, White DC, Peskett S. Lumbar epidural injections in the treatment of sciatica. Br. J. Rheumatol. Feb 1984;23(1):35-38.
107.
14 15
M AN U
11
Kim M-J, Lee S-H, Jung E-S, et al. Targeted percutaneous transforaminal
SC
105.
Koranda I, Sefrna F. Effectiveness of conservative and surgical treatment of lumboischiadic syndrome. Cas Lek Cesk. Aug 2 1995;134(15):474-477.
108.
Korhonen T, Karppinen J, Paimela L, et al. The treatment of disc herniation-induced
TE D
8
16
sciatica with infliximab - Results of a randomized, controlled, 3-month follow-up
17
study. Spine. Dec 2005;30(24):2724-2728.
18
109.
Kwasucki J, Olbrych K, ska B, Stepie A, Dec S, Maksymiuk G. Assessment of dexamethasone effectiveness in the treatment of ischalgia. Neurol Neurochir Pol.
20
1993;27(4):515-522. 110.
22
action of fluvoxamine compared with efficacy of imipramine and tramadol for
23 24
Kwasucki J, Stepien A, Maksymiuk G, Olbrych-Karpinska B. Evaluation of analgesic
AC C
21
EP
19
treatment of sciatica--open trial. Wiad Lek. 2002;55(1-2):42-50.
111.
Lagarrigue J, Lazorthes Y, Verdie JC, Richaud J. Comparative Results of Surgery
25
and Nucleolysis in 1085 Cases of Herniated Lumbar Disc. Neurochirurgie.
26
1991;37(2):96-105.
29
ACCEPTED MANUSCRIPT
1
112.
Laiq N, Khan MN, Iqbal MJ, Khan S. Comparison of Epidural Steroid Injections with
2
conservative management in patients with lumbar radiculopathy. J Coll Physicians
3
Surg Pak. 2009;19(9):539-543. 113.
Larsson U, Choler U, Lidstrom A, et al. Auto Traction for Treatment of Lumbago
RI PT
4 5
Sciatica a Multicenter Controlled Investigation. Acta Orthop Scand Suppl.
6
1980;51(5):791-798.
7
114.
Lavignolle B, Vital JM, Baulny D, Grenier F, Castagnera L. Comparative study of
surgery and chemonucleolysis in the treatment of sciatica caused by a herniated
9
disk. Acta Orthop Belg. 1987;53(2):244-249. 115.
Lee SH, Lee SJ, Park KH, et al. Comparison of percutaneous manual and
M AN U
10
SC
8
11
endoscopic laser diskectomy with chemonucleolysis and automated nucleotomy.
12
Orthopade. 1996;25(1):49-55.
13
116.
Lee SH, Chung SE, Ahn Y, Kim TH, Park JY, Shin SW. Comparative radiologic evaluation of percutaneous endoscopic lumbar discectomy and open
15
microdiscectomy: A matched cohort analysis. Mount Sinai Journal of Medicine. Sep
16
2006;73(5):795-801. 117.
18 19
Lidstrom A, Zachrisson M. Physical therapy on low back pain and sciatica. An attempt at evaluation. Scand J Rehabil Med. 1970;2(1):37-42.
118.
Ljunggren AE, Walker L, Weber H, Amundsen T. Manual traction versus isometric
EP
17
TE D
14
exercises in patients with herniated intervertebral lumbar discs. Physiotherapy
21
Theory and Practice. Dec 1992;8(4):207-213.
22
119.
23
Luijsterburg PAJ, Verhagen AP, Ostelo RWJG, et al. Physical therapy plus general practitioners' care versus general practitioners' care alone for sciatica: a randomised
24
clinical trial with a 12-month follow-up. European Spine Journal. Apr 2008;17(4):509-
25 26
AC C
20
517. 120.
Lundin A, Magnuson A, Axelsson K, Kogler H, Samuelsson L. The effect of
27
perioperative corticosteroids on the outcome of microscopic lumbar disc surgery.
28
European Spine Journal. Dec 2003;12(6):625-630.
30
ACCEPTED MANUSCRIPT
1
121.
Mackay MA, Fischgrund JS, Herkowitz HN, Kurz LT, Hecht B, Schwartz M. The
2
Effect of Interposition Membrane on the Outcome of Lumbar Laminectomy and
3
Diskectomy. Spine. Aug 1995;20(16):1793-1796. 122.
Mathews JA, Mills SB, Jenkins VM, et al. Back pain and sciatica: controlled trials of
RI PT
4 5
manipulation, traction, sclerosant and epidural injections. Br J Rheumatol.
6
1987;26(6):416-423.
7
123.
Mayer HM, Brock M. Percutaneous Endoscopic Discectomy - Surgical Technique and Preliminary-Results Compared to Microsurgical Discectomy. Journal of
9
Neurosurgery. Feb 1993;78(2):216-225. 124.
Moret NC, van der Stap M, Hagmeijer R, Molenaar A, Koes BW. Design and
M AN U
10
SC
8
11
feasibility of a randomized clinical trial to evaluate the effect of vertical traction in
12
patients with a lumbar radicular syndrome. Manual Therapy. 1998;3(4):203-211.
13
125.
Muralikuttan KP, Hamilton A, Kernohan WG, Mollan RA, Adair IV. A prospective randomized trial of chemonucleolysis and conventional disc surgery in single level
15
lumbar disc herniation. Spine. Apr 1992;17(4):381-387.
16
126.
TE D
14
Murata Y, Kato Y, Miyamoto K, Takahashi K. Clinical study of low back pain and
17
radicular pain pathways by using l2 spinal nerve root infiltration: a randomized,
18
controlled, clinical trial. Spine. 2009;34(19):2008-2013. 127.
North RB, Kidd DH, Farrokhi F, Piantadosi SA. Spinal cord stimulation versus
EP
19
repeated lumbosacral spine surgery for chronic pain: a randomized, controlled trial.
21
Neurosurgery. 2005;56(1):98-106; discussion 106-107.
22
128.
23 24
AC C
20
Norton WL. Chemonucleolysis versus surgical discectomy. Comparison of costs and results in workers' compensation claimants. Spine. Jun 1986;11(5):440-443.
129.
Osterman H, Seitsalo S, Karppinen J, Malmivaara A. Effectiveness of
25
microdiscectomy for lumbar disc herniation: a randomized controlled trial with 2 years
26
of follow-up. Spine. Oct 1 2006;31(21):2409-2414.
31
ACCEPTED MANUSCRIPT
1
130.
Ozturk B, Gunduz OH, Ozoran K, Bostanoglu S. Effect of continuous lumbar traction
2
on the size of herniated disc material in lumbar disc herniation. Rheumatology
3
International. May 2006;26(7):622-626. 131.
Peul WC, Van Houwelingen HC, Van Den Hout WB, et al. Surgery versus prolonged
RI PT
4 5
conservative treatment for sciatica. New England Journal of Medicine. 31
6
2007;356(22):2245-2256.
7
132.
Popiolek A, Domanik A, Mazurkiewicz G. Epidural injections of steroids in the
treatment of patients with chronic sciatica in discopathy. Neurol Neurochir Pol. Sep-
9
Oct 1991;25(5):640-646. 133.
Porsman O, Friis H. Prolapsed lumbar disc treated with intramuscularly administered
M AN U
10
SC
8
11
dexamethasonephosphate. A prospectively planned, double-blind, controlled clinical
12
trial in 52 patients. Scandinavian Journal of Rheumatology. 1979;8(3):142-144.
13
134.
Postacchini F, Lami R, Massobrio M. Chemonucleolysis versus surgery in lumbar disc herniations: correlation of the results to preoperative clinical pattern and size of
15
the herniation. Spine. Mar 1987;12(2):87-96.
16
135.
TE D
14
Price C, Arden N, Coglan L, Rogers P. Cost-effectiveness and safety of epidural
17
steroids in the management of sciatica. Health Technology Assessment. Aug
18
2005;9(33):1-58. 136.
Rasmussen S, Krum-Moller DS, Lauridsen LR, et al. Epidural steroid following
EP
19
discectomy for herniated lumbar disc reduces neurological impairment and enhances
21
recovery: a randomized study with two-year follow-up. Spine. 2008;33(19):2028-
22 23
AC C
20
2033.
137.
24
Rattanatharn R, Sanjaroensuttikul N, Anadirekkul P, Chaivisate R, Wannasetta W. Effectiveness of lumbar traction with routine conservative treatment in acute
25
herniated disc syndrome. Journal of the Medical Association of Thailand. Sep
26
2004;87 Suppl 2:S272-277.
27 28
138.
Reust P, Chantraine A, Vischer TL. Treatment of lumbar sciatica with or without neurological deficit using mechanical traction. A double-blind study. Schweizerische
32
ACCEPTED MANUSCRIPT
1
Medizinische Wochenschrift. Feb 27 1988;Journal Suisse de Medecine. 118(8):271-
2
274.
3
139.
Revel M, Payan C, Vallee C, et al. Automated Percutaneous Lumbar Discectomy Versus Chemonucleolysis in the Treatment of Sciatica - a Randomized Multicenter
5
Trial. Spine. Jan 1993;18(1):1-7.
6
140.
RI PT
4
Richter HP, Kast E, Tomczak R, Besenfelder W, Gaus W. Results of applying
ADCON-L gel after lumbar discectomy: the German ADCON-L study. Journal of
8
Neurosurgery. Oct 2001;95(2 Suppl):179-189.
9
141.
SC
7
Ridley MG, Kingsley GH, Gibson T, Grahame R. Outpatient lumbar epidural corticosteroid injection in the management of sciatica. Br. J. Rheumatol. Aug
11
1988;27(4):295-299.
12
142.
M AN U
10
Ronnberg K, Lind B, Zoega B, et al. Peridural scar and its relation to clinical
13
outcome: a randomised study on surgically treated lumbar disc herniation patients.
14
European Spine Journal. 2008;17(12):1714-1720. 143.
Santilli V, Beghi E, Finucci S. Chiropractic manipulation in the treatment of acute
TE D
15 16
back pain and sciatica with disc protrusion: a randomized double-blind clinical trial of
17
active and simulated spinal manipulations. Spine J. Mar-Apr 2006;6(2):131-137.
18
144.
Schwetschenau PR, Ramirez A, Johnston J. Double blind evaluation of intradiscal chymopapain for herniated lumbar discs; early results. J Neurosurg. 1976;45(6):622-
20
627. 145.
22
analysis of extended conservative therapy versus surgical intervention in the
23 24
Shvartzman L, Weingarten E, Sherry H, Levin S, Persaud A. Cost-effectiveness
AC C
21
EP
19
management of herniated lumbar intervertebral disc. Spine. Feb 1992;17(2):176-182.
146.
Snoek W, Weber H, Jorgensen B. Double-Blind Evaluation of Extradural Methyl
25
Prednisolone for Herniated Lumbar Discs. Acta Orthop Scand Suppl.
26
1977;48(6):635-641.
33
ACCEPTED MANUSCRIPT
1
147.
Steffen R. Laser discotomy/Chemonucleolysis ini the therapy of displaced lumbal
2
intervertebral disc - prospective randomised comparison. Hefte zur der Unfallchirurg.
3
1999:215-221. 148.
Stula D. Chemonucleolysis with Chymopapain in Intervertebral-Disk Hernia - a
RI PT
4 5
Randomized Comparative-Study of Patients Following Surgery. Neurochirurgia
6
(Stuttg). Sep 1990;33(5):169-172.
7
149.
Styczynski T, Zarski S, Drozdowska Z, et al. Clinical Evaluation of the Effectiveness of Traction Treatment of Patients with Herniation of Lumbar Intervertebral Discs.
9
Reumatologia (Warsaw). 1991;29(3-4):275-281. 150.
Tassi GP. Comparison of results of 500 microdiscectomies and 500 percutaneous
M AN U
10
SC
8
11
laser disc decompression procedures for lumbar disc herniation. Photomed Laser
12
Surg. Dec 2006;24(6):694-697.
13
151.
Tregonning GD, Transfeldt EE, McCulloch JA, Macnab I, Nachemson A. Chymopapain versus conventional surgery for lumbar disc herniation. 10-year results
15
of treatment. J Bone Joint Surg Br. May 1991;73(3):481-486.
16
152.
TE D
14
de Tribolet N, Porchet F, Lutz TW, et al. Clinical assessment of a novel antiadhesion barrier gel: prospective, randomized, multicenter, clinical trial of ADCON-L to inhibit
18
postoperative peridural fibrosis and related symptoms after lumbar discectomy. Am J
19
Orthop. 1998;27(2):111-120. 153.
21
modalities for acute pain in lumbar disc herniation measured by clinical evaluation
22
and magnetic resonance imaging. J Manipulative Physiol Ther. Mar 2008;31(3):191-
23 24
Unlu Z, Tasci S, Tarhan S, Pabuscu Y, Islak S. Comparison of 3 physical therapy
AC C
20
EP
17
198.
154.
Vad VB, Bhat AL, Lutz GE, Cammisa F. Transforaminal epidural steroid injections in
25
lumbosacral radiculopathy: a prospective randomized study. Spine. Jan 1
26
2002;27(1):11-16.
34
ACCEPTED MANUSCRIPT
1
155.
Valat JP, Giraudeau B, Rozenberg S, et al. Epidural corticosteroid injections for
2
sciatica: a randomised, double blind, controlled clinical trial. Ann Rheum Dis. Jul
3
2003;62(7):639-643. 156.
van Alphen HA, Braakman R, Bezemer PD, Broere G, Berfelo MW.
RI PT
4 5
Chemonucleolysis versus discectomy: a randomized multicenter trial. Journal of
6
Neurosurgery. Jun 1989;70(6):869-875.
7
157.
Veihelmann A, Devens C, Trouillier H, Birkenmaier C, Gerdesmeyer L, Refior HJ.
Epidural neuroplasty versus physiotherapy to relieve pain in patients with sciatica: a
9
prospective randomized blinded clinical trial. J Orthop Sci. Jul 2006;11(4):365-369. 158.
Vroomen PC, de Krom MC, Wilmink JT, Kester AD, Knottnerus JA. Lack of
M AN U
10
SC
8
11
effectiveness of bed rest for sciatica. New England Journal of Medicine. Feb 11
12
1999;340(6):418-423.
13
159.
Watters WC, Mirkovic S, Boss J. Treatment of the isolated lumbar intervertebral disc herniation: microdiscectomy versus chemonucleolysis. Spine (Phila Pa 1976).
15
1988;13(3):360-362.
16
160.
TE D
14
Watts C, Hutchison G, Stern J, Clark K. Comparison of intervertebral disc disease
17
treatment by chymopapain injection and open surgery. Journal of Neurosurgery. Apr
18
1975;42(4):397-400.
20 21
observation. Spine. Mar 1983;8(2):131-140. 162.
22
Weber H, Holme I, Amlie E. The natural course of acute sciatica with nerve root symptoms in a double-blind placebo-controlled trial evaluating the effect of piroxicam.
23 24
Weber H. Lumbar disc herniation. A controlled, prospective study with ten years of
EP
161.
AC C
19
Spine. Sep 1 1993;18(11):1433-1438.
163.
Wehling P, Reinecke J. Acupuncture together with cytokine depressing herbs in
25
comparison to injection therapy with steroids in sciatic pain. Schmerz. Jun 13
26
1997;11(3):180-184.
27 28
164.
Weinstein JN, Tosteson TD, Lurie JD, et al. Surgical vs nonoperative treatment for lumbar disk herniation: the Spine Patient Outcomes Research Trial (SPORT): a
35
ACCEPTED MANUSCRIPT
1
randomized trial. Journal of the American Medical Association. Nov 22
2
2006;296(20):2441-2450.
3
165.
Weinstein JN, Lurie JD, Tosteson TD, et al. Surgical vs nonoperative treatment for lumbar disk herniation: The Spine Patient Outcomes Research Trial (SPORT)
5
observational cohort. Journal of the American Medical Association. 22
6
2006;296(20):2451-2459.
7
166.
RI PT
4
Weinstein J, Spratt KF, Lehmann T, McNeill T, Hejna W. Lumbar disc herniation. A
comparison of the results of chemonucleolysis and open discectomy after ten years.
9
J Bone Joint Surg Am. Jan 1986;68(1):43-54. 167.
11 12
with chronic radiculopathy. The Pain Clinic. 2003;15(3):213-218. 168.
13 14
Yildirim K, Sisecioglu M, Karatay S, et al. The effectiveness of gabapentin in patients
M AN U
10
SC
8
Zeiger HE, Jr. Comparison of chemonucleolysis and microsurgical discectomy for the treatment of herniated lumbar disc. Spine. Oct 1987;12(8):796-799.
169.
Armon C, Argoff CE, Samuels J, Backonja MM. Assessment: Use of epidural steroid injections to treat radicular lumbosacral pain: Report of the Therapeutics and
16
Technology Assessment Subcommittee of the American Academy of Neurology.
17
Neurology. Mar 2007;68(10):723-729.
18
170.
TE D
15
Chou R, Atlas SJ, Stanos SP, Rosenquist RW. Nonsurgical interventional therapies for low back pain: a review of the evidence for an American Pain Society clinical
20
practice guideline. Spine 2009;34(10):1078-1093. 171.
22 23
studies. Health Technol Assess. 2003;7(27).
172.
24 25 26
Deeks JJ, Dinnes J, D'Amico R, et al. Evaluating non-randomised intervention
AC C
21
EP
19
Turner RM, Spiegelhalter DJ, Smith GCS, Thompson SG. Bias modelling in evidence synthesis. J R Stat Soc Ser A Stat Soc. 2009;172(19381328):21-47.
173.
Eddy DM, Hasselblad V, Shachter R. An introduction to Bayesian methods for Metaanalysis: The Confidence Profile Method. Medical Decision Making. 1990;10:15-23.
36
ACCEPTED MANUSCRIPT
1
174.
Egger M, Smith GD, Altman D, Schneider M. Chapter 12: Systematic reviews of
2
observational studies. In: Egger M, Smith GD, Altman D, eds. Systematic Reviews in
3
Health Care: Meta-analysis in Context. London: BMJ Books; 2001. 175.
Wiebe N, Vandermeer B, Platt RW, Klassen TP, Moher D, Barrowman NJ. A
RI PT
4 5
systematic review identifies a lack of standardization in methods for handling missing
6
variance data. J Clin Epidemiol. Apr 2006;59(4):342-353. 176.
8 9
Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions. Vol Version 5.0.2. 2 ed. Chichester (UK): John Wiley & Sons Ltd; 2009.
177.
SC
7
Thompson S, Ekelund U, Jebb S, et al. A proposed method of bias adjustment for meta-analyses of published observational studies. International Journal of
11
Epidemiology 2011;40:765–777
12
M AN U
10
Figure legends:
14
Figure 1: flow diagram showing the number of references identified, publications retrieved for
15
assessment, and studies included in the review
16
Figure 2: Network of treatment strategies for sciatica for comparative studies reporting
17
global effect
18
Figure 3: Network of treatment strategies for sciatica for comparative studies reporting pain
19
intensity
20
Figure 4: Plot of the odds ratios (ORs) of global effect for different treatment strategies
21
compared with inactive control from the network meta-analysis
22
Figure 5: Plot of the weighted mean difference for pain intensity for different treatment
23
strategies compared with inactive control from the network meta-analysis
AC C
EP
TE D
13
37
ACCEPTED MANUSCRIPT
SUPPLEMENTARY DATA Manuscript title: Comparative clinical effectiveness of management strategies for sciatica: systematic review and mixed treatment comparisons metaanalyses.
RI PT
INCLUDES:
Web appendix A (pg 3): Winbug codes.
design and outcome data for the pairwise studies. Table B1 (pg 6): Description of interventions.
M AN U
Table B2 (pg 11): Description of participants.
SC
Web appendix B (pg 6): Description of the interventions, population, study
Table B3 (pg 21): Summary of overall quality of included studies. Table B4 (pg 25): Outcome data for global effect.
Table B5 (pg 33): Outcome data for pain intensity.
Web appendix C (pg 37): Results of inactive control comparisons from network meta-analyses.
TE D
Table C1 (pg 37): Probability of being best and the ORs of global effect for different treatment strategies compared with inactive control from the network meta-analysis and pairwise meta-analyses.
Table C2 (pg 38): Probability of being best and the ORs for global effect of different treatment strategies compared with inactive control from the network meta-analysis
EP
and pairwise meta-analyses based on RCTs and Q-RCTs only. Figure C1 (pg 39): Plot of the ORs of global effects for the different treatment
AC C
strategies compared with inactive control from the network meta-analysis, based on RCTs and Q-RCTs only. Table C3 (pg 40): Probability of being best and the weighted mean difference for pain intensity for different treatment strategies compared with inactive control from the network meta-analysis and pairwise meta-analyses. Table C4 (pg 41): Probability of being best and weighted mean difference for pain intensity of different treatment strategies compared with inactive control from the network meta-analysis and pairwise meta-analyses based on RCTs and Q-RCTs only.
1
ACCEPTED MANUSCRIPT
Figure C2 (pg 42): Plot of the weighted mean difference for pain intensity for the different treatment strategies compared with inactive control from the MTC analysis, based on RCTs and Q-RCTs only.
Web appendix D (pg 43): Results of network meta-analyses restricted to RCTs
RI PT
and Q-RCTs.
Table D1 (pg 43): Results (odds ratios ORs, with 95% confidence intervals/credible intervals) of network meta-analysis for RCTs/Q-RCTs reporting global effect.
Table D2 (pg 45): Results (weighted mean difference WMD, with 95% confidence intervals/credible intervals) of network meta-analysis for RCTs/Q-RCTs reporting
SC
pain intensity.
M AN U
Results of sensitivity analyses (pg 46) (a narrative summary)
AC C
EP
TE D
Appendix E (pg 47): Assessment of model fit and between study heterogeneity
2
ACCEPTED MANUSCRIPT
WEB APPENDIX A: WINBUG CODES These are based on those published on the Bristol University network meta-analysis webpage (https://www.bris.ac.uk/cobm/research/mpes/mtc.html)
M AN U
SC
RI PT
GLOBAL #Random-effects model for multi-arm trials model{ for(i in 1:NS){ w[i,1] 1.0 favours intervention compared with control.
43
U
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intra-operative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; S intradiscal injections; T Spinal cord simulations; U Radiofrequency treatment ; N number of studies included in conventional pairwise meta-analysis.
44
ACCEPTED MANUSCRIPT
Table D2: Results (weighted mean difference WMD, with 95% confidence intervals/credible intervals) of network meta-analysis for
N=1, 6.0 (-4, 16)
N=1, -2.00 (-12, 8) N=1, 9.00 (-4, 22)
N=8,-5.17 (-12, 2) N=2,18.01 (6, 30)
D 4.56 (-18, 28) 6.49 (-5, 17) -6.74 (-30, 17) 8.64 (-13, 31) 4.06 (-31, 39) -14.55 (-44, 15) 5.39 (-19, 30) 9.93 (-10, 30) 2.80 (-15, 21) 27.64 (-5, 61) 14.43 (-8, 36) 26.83 (-10, 65) 22.64 (-11, 56) -1.79 (-23, 20) 23.39 (-3, 50)
N=1, -7.00 (-14, -0.4)
N=1, 35.00 (-25, 95) N=1, -0.63 (-15, 14)
E 1.84 (-23, 26) -11.36 (-34, 11) 4.10 (-26, 35) -0.5 (-27, 26) -19.04 (-56, 17) 0.76 (-26, 28) 5.38 (-23, 35) -1.73 (-30, 27) 23.14 (-16, 63) 9.80 (-21, 40) 22.23 (-20, 66) 18.02 (-6, 43) -6.39 (-37, 24) 18.69 (-15, 53)
F -13.23 (-38, 12) 2.19 (-20, 25) -2.39 (-38, 34) -21.0 (-51, 9) -1.12 (-28, 26) 3.41 (-17, 25) -3.69 (-23, 17) 21.22 (-12, 56) 7.97 (-12, 28) 20.3 (-17, 59) 16.11 (-18, 51) -8.28 (-29, 13) 16.86 (-10, 44)
G 15.40 (-15, 47) 10.8 (-24, 46) -7.752 (-45, 29) 12.16 (-10, 34) 16.68 (-13, 47) 9.62 (-19, 39) 34.56 (-5, 75) 21.2 (-10, 52) 33.6 (-10, 77) 29.37 (-4, 63) 4.92 (-26, 36) 30.11 (-5, 65)
-4.65 (-45, 35) -23.2 (-58, 11) -3.30 (-36, 29) 1.31 (-15, 17) -5.85 (-33, 21) 18.99 (-6, 44) 5.78 (-24, 34) 18.15 (-12, 49) 13.93 (-25, 53) -10.38 (-40, 18) 14.66 (-18, 47)
I -18.67 (-64, 26) 1.25 (-37, 39) 5.90 (-33, 46) -1.27 (-40, 38) 23.61 (-24, 72) 10.36 (-31, 50) 22.64 (-27, 74) 18.54 (-17, 55) -5.89 (-46, 34) 19.14 (-24, 62)
N=1,13.00 (2, 24)
N=1, -9.3 (-23, 4.9)
N=1, 18.00 (8, 28) N=1, 22.50 (11, 35)
N=2, 3.19 (-13, 19)
H
N=1, -26.66 (-38, -15)
N=1, -7.0 (-5, 19)
SC
C 1.65 (-20, 23) 6.23 (-14, 27) 8.12 (-16, 32) -5.09 (-14, 4) 10.3 (-19, 40) 5.75 (-28, 40) -12.81 (-49, 23) 7.04 (-13, 27) 11.61 (-17, 41) 4.47 (-23, 32) 29.48 (-9, 69) 16.11 (-14, 46) 28.5 (-14, 71) 24.26 (-8, 57) -0.16 (-30, 30) 24.99 (-9, 59)
N=1, -25.00 (-42, -8)
N= 1, 3.36 (-15, 21)
M AN U
N=1, -6.1 (-11, -0.8)
N=4, -6.26 (-15, 3)
N=1, 19.00 (8, 30)
J
19.86 (-18, 58) 24.48 (-9, 59) 17.37 (-16, 51) 42.28 (-0.1, 85) 28.89 (-6, 64) 41.4 (-4, 88) 37.07 (-6, 81) 12.74 (-22, 47) 37.93 (0.2, 76)
TE D
B -7.35 (-25, 10) -5.70 (-22, 11) -1.07 (-24, 22) 0.81 (-19, 20) -12.46 (-32, 7) 2.99 (-23, 30) -1.56 (-37, 34) -20.16 (-54, 13) -0.33 (-20, 20) 4.31 (-21, 30) -2.86 (-27, 21) 22.01 (-14, 59) 8.78 (-19, 35) 21.17 (-19, 62) 16.97 (-17, 51) -7.49 (-34, 19) 17.62 (-13, 48)
N=1, -5.40 (-24, 13)
EP
-4.61 (-22, 13) -11.96 (-34, 10) -10.3 (-19, -2) -5.68 (-28, 17) -3.82 (-14, 6) -17.04 (-41, 7) -1.64 (-21, 19) -6.26 (-41, 29) -24.8 (-53, 3) -4.95 (-30, 20) -0.36 (-18, 19) -7.49 (-25, 11) 17.35 (-14, 50) 4.19 (-17, 25) 16.51 (-19, 53) 12.26 (-21, 46) -12.11 (-32, 8) 13.03 (-12, 38)
N=7, -8.11 (-19, 3) N=2, -5.32 (-12, 1)
AC C
A
RI PT
RCTs/Q-RCTs reporting pain intensity
K
4.60 (-26, 36) -2.57 (-32, 28) 22.38 (-18, 64) 9.04 (-24, 41) 21.5 (-22, 66) 17.26 (-19, 54) -7.27 (-40, 25) 17.97 (-18 54)
L
-7.11 (-33, 18) 17.73 (-12, 48) 4.51 (-24, 32) 16.81 (-18, 51) 12.66 (-26, 50) -11.73 (-40, 15) 13.36 (-18, 44)
M 24.8 (-12, 62) 11.62 (-16, 39) 23.98 (-16, 65) 19.84 (-18, 57) -4.59 (-32, 22) 20.54 (-11, 51)
N=2, -1.09 (-7, 5)
N -13.24 (-53, 24) -0.85 (-18, 16) -5.13 (-52, 41) -29.52 (-68, 8) -4.301 (-45, 36)
N=2, -8.23 (-18, 2)
O 12.46 (-29, 55) 8.17 (-31, 48) -16.23 (-33, 1) 8.86 (-24, 42)
P -4.21 (-54, 45) -28.63 (-71, 12) -3.49 (-48, 40)
Q -24.45 (-64, 15) 0.70 (-41, 42)
R 25.09 (-7, 57)
Lower triangle includes the findings of the network meta-analysis (posterior median weighted mean differences WMDs plus 95% credible intervals) conducted in the Bayesian statistical package WinBUGS; upper triangle includes the findings of the direct standard pairwise meta-analyses (WMD plus confidence intervals) conducted using STATA Statistically significant findings have been shaded (significance assessment made on data rounded to decimal places) WMD > 0 (representing reduction in pain) favours intervention compared with control A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intra-operative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; U Radiofrequency treatment; N number of studies included in conventional pairwise meta-analysis.
45
U
ACCEPTED MANUSCRIPT
Results of sensitivity analyses
The results of the sensitivity analyses excluding observational studies and non-RCTs showed broad agreement with the main analyses. For global effect, the most notable discrepancies occurred with biological agents compared with chemonucleolysis,
RI PT
conventional care, and exercise therapy. The relative effects of epidural, chemonucleolysis, non-opioids, manipulations, and acupuncture with usual care were no longer statistically
significant in the network meta-analysis restricted to RCTs and Q-RCTs. A number of other
comparisons were also no longer statistically significant, usually with the analyses restricted to RCTs having a wider credible interval (epidural injections vs chemonucleolysis, traction,
SC
exercise therapy, and spinal cord simulation SCS; and SCS vs non-opioids, manipulation, and accupuncture). The following comparison, however, did become statistically significant
M AN U
within the restricted analyses: disc surgery was more effective than radiofrequency treatment (OR 6.68, 95% CrI: 1.19 to 39.53); intra-operative injections were more effective than both non-opioids (OR 3.16, 95% CrI: 1.09 to 9.49), passive physical therapy (OR 3.89, 95% CrI: 1.11 to 13.84), opioids (OR 5.40, 95% CrI: 1.35 to 22.40), and neuropathic pain modulators (OR 4.33, 95% CrI: 1.09 to 17.51); and manipulation was more effective radiofrequency treatment (OR 8.84, 95% CrI: 1.06 to 76.22).
TE D
For pain intensity the most notable discrepancies between the network meta-analysis with and without observational studies and non-RCTs only occurred with biological agents (vs inactive control, conventional care, disc surgery, non-opioids, intra-operative interventions, acupuncture, exercise therapy, opioids, and neuropathic painmodulators). Only the mean
EP
difference between epidural injections and inactive control (WMD -10.3 95% CrI: -18.5, 2.24) remained statistically significant in the network meta-analysis restricted to RCTs and Q-RCTs, predominantly through the inflation of confidence intervals due to the removal of
AC C
evidence.
46
ACCEPTED MANUSCRIPT
Appendix E: Assessment of model fit and between study heterogeneity MODEL FIT Table E: Residual deviance data for all network meta-analyses No. of data points 190 136 107 92
Posterior mean deviance 195 140 105 90
RI PT
Model Global effect including all studies Global effect including RCTs and Q-RCTs Pain intensity including all studies Pain intensity including RCTs and Q-RCTs
SC
HETEROGENEITY Global effect
M AN U
For the network meta-analyses of global effect that included all study designs, the median between-trial variance (tau-squared) observed in the posterior distributions was 0.38 (95% CrI: 0.22 to 0.65). However, this does not give an indication of the between study heterogeneity within each intervention comparison. This information has been derived from the standard pair-wise meta-analyses shown in the Table E1 below. There was high a level of between study heterogeneity for the following comparisons: disc surgery vs conventional care; and percutaneous discectomy vs disc surgery. Table E1: Test(s) of heterogeneity for studies included in the standard pair-wise analyses for global effect which included all study types
TE D
Degrees of freedom 4 8 2 4 13 6 0 0 6 3 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
EP
Chi-squared statistic 10.5 26.58 8.82 5.01 41.08 7.74 0.00 0.00 2.32 13.10 0.18 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.76
AC C
Treatment comparators CB DA DB EA EC FA FC FD GC GE HA HK IA JF KB KC KH LA LD LH MA ND NH NK OA OF PN
P-value
I-squared**
Tau-squared
0.038 0.001 0.012 0.286 0.000 0.258 . . 0.888 0.004 0.669 . . . . . . . . . . . . . . . 0.384
60.6% 69.9% 77.3% 20.2% 68.4% 22.5% .% .% 0.0% 77.1% 0.0% .% .% .% .% .% .% .% .% .% .% .% .% .% .% .% 0.0%
0.1806 0.7921 2.2995 0.0604 0.2582 0.0659 0.0659 0.0659 0.0000 0.7505 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
47
ACCEPTED MANUSCRIPT
I-squared**
Tau-squared
74.1% 49.0% 37.7% 0.0% 57.0% .% .% 76.3%
0.3051 0.1483 0.1814 0.0000 0.3459 0.3459 0.3459 0.5917
RI PT
Treatment Chi-squared Degrees of P-value comparators statistic freedom QC 15.43 4 0.004 QE 7.85 4 0.097 RA 1.61 1 0.205 RO 0.29 1 0.593 SE 6.97 3 0.073 TC 0.00 0 . UA 0.00 0 . Overall 396.34 94 0.000 ** I-squared: the variation in OR attributable to heterogeneity
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intraoperative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; S intradiscal injections; T Spinal cord simulations; U Radio frequency treatment.
M AN U
SC
For the network meta-analyses of global effect that included only the RCTs and quasiRCTs, the median between-trial variance (tau-squared) observed in the posterior distributions was 0.30 (95% CrI: 0.12 to 0.62). There was high a level of between study heterogeneity for the following comparisons: disc surgery vs conventional care; and percutaneous discectomy vs disc surgery (see Table E2). Table E2: Test(s) of heterogeneity for studies included in the standard pair-wise analyses for global effect which included only RCTs or quasi-RCTs
AC C
EP
TE D
Treatment Chi-squared Degrees of P-value comparators statistic freedom CB 8.96 2 0.011 DA 26.58 8 0.001 DB 0.17 1 0.680 EA 5.01 4 0.286 EC 7.26 4 0.123 FA 7.74 6 0.258 FD 0.00 0 . GC 2.32 6 0.888 HA 0.18 1 0.669 KH 0.00 0 . IA 0.00 0 . JF 0.00 0 . KB 0.00 0 . KC 0.00 0 . LA 0.00 0 . LD 0.00 0 . LH 0.00 0 . MA 0.00 0 . NH 0.44 1 0.506 OA 0.00 0 . OF 0.00 0 . PN 0.76 1 0.384 QC 7.25 1 0.007 QE 0.08 1 0.783 RA 1.61 1 0.205 RO 0.29 1 0.593 SE 5.13 2 0.077 TC 0.00 0 . UA 0.00 0 . Overall 201.29 68 0.000 ** I-squared: the variation in OR attributable to heterogeneity
I-squared**
Tau-squared
77.7% 69.9% 0.0% 20.2% 44.9% 22.5% .% 0.0% 0.0% .% .% .% .% .% .% .% .% .% 0.0% .% .% 0.0% 86.2% 0.0% 37.7% 0.0% 61.0% .% .% 66.2%
0.4647 0.7921 0.0000 0.0604 0.1757 0.0659 2.2417 0.0000 0.0000 0.3077 0.0000 0.3077 0.0000 0.0000 0.0000 2.2417 0.3077 0.0000 0.0000 0.0000 0.3077 0.0000 2.2417 0.0000 0.1814 0.0000 0.3077 2.2417 0.1814 0.4619
48
ACCEPTED MANUSCRIPT
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intraoperative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; S intradiscal injections; T Spinal cord simulations; U Radiofrequency treatment.
RI PT
Pain intensity
SC
For the network meta-analyses of pain intensity that included all study designs, the median between-trial variance (tau-squared) observed in the posterior distributions was 119.9 (95% CrI: 64.34 to 227.9). There was high a level of between study heterogeneity for the following comparisons: disc surgery vs inactive control; non-opioids vs inactive control; biological agents vs inactive control; intra-operative interventions vs disc surgery; and non-opioids vs epidural injections (see Table E3). Table E3: Test(s) of heterogeneity for studies included in the standard pair-wise analyses for pain intensity which included all study types P
I-squared**
M AN U
Treatment Heterogeneity degrees of comparator Statistic freedom (number of studies) DA 56.13 6 EA 0.00 0 FA 18.77 3 HA 0.00 0 JA 0.00 0 LA 0.00 0 MA 13.69 1 RA 0.00 0 UA 0.00 0 CB 1.02 1 DB 0.03 1 KB 0.00 0 EC 6.20 2 FC 0.00 0 GC 29.82 7 KC 0.00 0 FD 7.92 1 JD 0.00 0 LD 0.00 0 IE 0.00 0 QE 0.00 0 MF 0.00 0 OF 0.00 0 LH 3.86 1 NH 0.00 0 PN 0.16 1 RO 0.67 1 ** I-squared: the variation in WMD attributable to heterogeneity
AC C
EP
TE D
0.000 . 0.000 . . . 0.000 . . 0.313 0.868 . 0.045 . 0.000 . 0.005 . . . . . . 0.049 . 0.689 0.414
89.3% .% 84.0% .% .% .% 92.7% .% .% 1.8% 0.0% .% 67.7% .% 76.5% .% 87.4% .% .% .% .% .% .% 74.1% .% 0.0% 0.0%
Tau-squared
198.6672 0.0000 65.8051 0.0000 0.0000 0.0000 535.7751 0.0000 0.0000 0.7251 0.0000 0.0000 56.4674 0.0000 72.3876 0.0000 62.9147 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 98.4535 0.0000 0.0000 0.0000
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intraoperative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; U Radiofrequency treatment.
For the network meta-analyses of pain intensity that included only the RCTs and quasiRCTs, the median between-trial variance (tau-squared) observed in the posterior distributions was 122.4 (95% CrI: 62.15 to 249.9). There was high a level of between study heterogeneity for the following comparisons: disc surgery vs inactive control; non-opioids vs inactive control; biological agents vs inactive control; intra-operative interventions vs disc surgery; and non-opioids vs epidural injections (see Table E4).
49
ACCEPTED MANUSCRIPT
Table E4: Test(s) of heterogeneity for studies included in the standard pair-wise analyses for pain intensity which included only RCTs or quasi-RCTs I-squared**
Tau-squared
0.000 . 0.000 . . . 0.000 . . . 0.868 . . 0.000 . 0.005 . . . . 0.049 . 0.689 0.414
89.3% .% 84.0% .% .% .% 92.7% .% .% .% 0.0% .% .% 76.5% .% 87.4% .% .% .% .% 74.1% .% 0.0% 0.0%
198.6672 0.0000 65.8051 0.0000 0.0000 0.0000 535.7751 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 72.3876 0.0000 62.9147 0.0000 0.0000 0.0000 0.0000 98.4535 0.0000 0.0000 0.0000
SC
RI PT
P
M AN U
Treatment Heterogeneity degrees of comparator Statistic freedom (number of studies) DA 56.13 6 EA 0.00 0 FA 18.77 3 HA 0.00 0 JA 0.00 0 LA 0.00 0 MA 13.69 1 RA 0.00 0 UA 0.00 0 CB 0.00 0 DB 0.03 1 KB 0.00 0 EC 0.00 0 GC 29.82 7 KC 0.00 0 FD 7.92 1 LD 0.00 0 IE 0.00 0 QE 0.00 0 OF 0.00 0 LH 3.86 1 NH 0.00 0 PN 0.16 1 RO 0.67 1 ** I-squared: the variation in WMD attributable to heterogeneity
AC C
EP
TE D
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intraoperative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; U Radiofrequency treatment.
50
ACCEPTED MANUSCRIPT
Table 1: Treatment categorisation
Treatment strategy
Treatment strategy Codea
Inactive control
A
Type of treatment Placebo in any type or format – tablet, injection, epidural etc.
RI PT
a
Sham treatment in any format No treatment B
Conservative therapy
SC
Conventional care
Conventional care
Non-surgical treatments
Disc surgery
C
M AN U
General practitioner care Discectomy Micro-discectomy Laminectomy Hemilaminectomy
TE D
Surgical decompression
D
AC C
EP
Epidural injections (includes spinal nerve block)
Chemonucleolysis
Non-opioids
Lumbar epidural injection Transforminal epidural injection Intraforaminal injection Interlaminar epidural injection Caudal epidural injection Periradicular injection Spinal nerve root block
E
Chymopapain Collagenase Ozone
F
Conventional pain or anti-inflammatory medication used as oral, IV or intramuscular: Steroids
ACCEPTED MANUSCRIPT
Treatment a strategy Code
a
Treatment strategy
Type of treatment Non-steroidal anti-inflammatory drugs (NSAIDs) Paracetamol
G
Barrier membranes:
RI PT
Intra-operative interventions
-
Anti-adhesion barrier
-
Fat graft
Intra-operative steroid +/- LA H
Mechanical traction
SC
Traction
Antigravitational traction
M AN U
Auto-traction
Manual traction
Manipulation
I
Chiropractic Osteopathic
J
Exercise therapy
K
L
AC C
EP
Passive physical therapy
TE D
Acupuncture
Biological agents
Acupuncture Exercise therapy Isometric exercises Mobilising and strengthening exercises Ultrasound Transcutaneous electrical nerve stimulation (TENS) Infra-red heat Physical therapy programme (hot pack, continuous ultrasound, and diadynamic currents) Conservative physiotherapy
M
Cytokine modulating treatments targeting tumour necrosis factor alpha: - Entanercept - Infliximab - Autologous Conditioned Serum
ACCEPTED MANUSCRIPT
Treatment strategy
Treatment a strategy Code
Bed rest
N
Opioids
O
Oral, IV or intramuscular opioids
Education/Advice
P
Advice to keep active
Type of treatment
RI PT
a
Advised to continue activities of daily living Percutaneous discectomy
Q
Automated percutaneous discectomy
SC
Percutaneous automated nucleotomy Nucleoplasty
Neuropathic painmodulators
R
M AN U
Lasar discectomy
Pharmaceutical treatment used for neuropathic pain: Anti-epileptic medication Tricyclic antidepressants
S
Spinal cord stimulation
T
Radiofrequency treatment a
TE D
Intradiscal injections
U
Interventions are summarised using these codes for displaying the results of the network-meta-analyses
AC C
EP
IV intra-venous
ACCEPTED MANUSCRIPT
D 0.45 (0.2, 0.9) 0.63 (0.3, 1.2) 1.05 (0.4, 2.4) 0.37 (0.2, 0.8) 1.4 (0.3, 7.0) 2.27 (0.3, 19.9) 0.31 (0.1, 0.9) 0.43 (0.2, 1.1) 4.8 (0.2, 278) 0.39 (0.1, 1.3) 0.33 (0.1, 1.1) 0.50 (0.1, 2.5) 0.35 (0.2, 0.8) 0.4 (0.1, 1.3) 0.2 (0.1, 0.6) 0.80 (0.1, 5.1) 0.16 (0.03, 0.9)
N=1, 6.71 (0.8, 58.8) N=1, 0.45 (0.2, 1.4)
N=1, 4.27 (1.5, 12.4)
N=7, 1.49 (1, 2.2)
N=1, 0.2 (0.1, 0.6)
N=1, 3.27 (0.8, 13.9)
F 1.67 (0.7, 4.1) 0.59 (0.2, 1.5) 2.24 (0.4, 11.4) 3.62 (0.5, 28.4) 0.5 (0.2, 1.5) 0.68 (0.2, 2) 7.71 (0.3, 449) 0.62 (0.2, 2.4) 0.53 (0.2, 1.6) 0.80 (0.1, 4.4) 0.56 (0.2, 1.4) 0.65 (0.2, 2) 0.32 (0.1, 0.97) 1.30 (0.2, 8.4) 0.25 (0.04, 1.5)
N=1, 10 (0.7,166.7)
N=1, 1.46 (0.7, 2) N=1, 0.77 (0.18, 3.2)
N=4, 2.03 (0.8, 5.4)
E 1.39 (0.7, 3.0) 2.32 (1.4, 4.0) 0.8 (0.3, 1.9) 3.1 (0.6, 16.3) 5.0 (0.7, 45.5) 0.7 (0.3, 1.9) 0.95 (0.3, 2.8) 10.77 (0.5, 625) 0.86 (0.2, 3.2) 0.73 (0.2, 2.5) 1.12 (0.2, 6.0) 0.78 (0.5, 1.3) 0.89 (0.3, 3.0) 0.44 (0.2, 1) 1.80 (0.3, 10.2) 0.35 (0.1, 2.1)
N=1, 4.72 (1.9, 11.4)
G 0.35 (0.1, 0.9) 1.34 (0.2, 7.5) 2.18 (0.3, 20.7) 0.30 (0.1, 0.9) 0.41 (0.1, 1.3) 4.65 (0.2, 277) 0.37 (0.1, 1.5) 0.32 (0.1, 1.2) 0.48 (0.1, 2.7) 0.34 (0.2, 0.7) 0.39 (0.1, 1.4) 0.19 (0.1, 0.5) 0.77 (0.1, 4.6) 0.15 (0.02, 0.96)
N=1, 0.87 (0.3, 2.7)
H 3.83 (0.7, 21.4) 6.21 (0.8, 59.0) 0.86 (0.3, 2.4) 1.17 (0.4, 3.2) 13.3 (0.5, 792) 1.06 (0.3, 3.8) 0.9 (0.2, 3.4) 1.37 (0.3, 7.2) 0.96 (0.4, 2.5) 1.1 (0.3, 4.1) 0.54 (0.2, 1.8) 2.18 (0.3, 14.8) 0.43 (0.1, 2.7)
I 1.63 (0.1, 22.3) 0.22 (0.04, 1.4) 0.3 (0.05, 1.8) 3.48 (0.1, 252) 0.28 (0.04, 2.0) 0.24 (0.04, 1.5) 0.36 (0.04, 3.4) 0.25 (0.04, 1.4) 0.29 (0.04, 1.8) 0.14 (0.02, 0.9 0.57 (0.1, 6.06) 0.11 (0.00, 1.1)
J 0.14 (0.01, 1.2) 0.19 (0.02, 1.7) 2.15 (0.05, 185) 0.17 (0.01, 1.7) 0.14 (0.01, 1.3) 0.22 (0.02, 2.8) 0.15 (0.02, 1.2) 0.18 (0.02, 1.6) 0.09 (0.01, 0.8) 0.35 (0.02, 5.1) 0.07 (0.01, 0.9)
N=1, 1.37 (0.5,3.8)
N=1, 0.93 (0.5, 1.9)
N=1, 0.57 (0.2, 1.7)
N=5, 0.67 (0.4, 1.3)
N=1, 1.13 (0.4, 3.6)
N=5, 0.65 (0.4, 1.1)
N=4, 0.46 (0.2, 1)
N=1, 0.18 (0.1, 0.7)
N=1, 1 (0.1, 7.1)
N=1, 2.2 (0.6, 7.6)
K
1.37 (0.4, 5.0) 15.63 (0.6, 977) 1.24 (0.4, 4.3) 1.05 (0.2, 4.7) 1.6 (0.3, 8.3) 1.13 (0.4, 3.3) 1.29 (0.3, 5.6) 0.64 (0.2, 2.3) 2.56 (0.4, 18.5 0.51 (0.1, 3.6)
N=2, 2.01 (0.8, 5.2)
N=1, 0.22 (0.1, 0.9)
SC
N=14, 0.5 (0.4, 0.7)
C 1.42 (0.7, 2.9) 0.64 (0.5, 0.9) 0.89 (0.4, 1.9) 1.48 (0.9, 2.5) 0.52 (0.2, 1.2) 1.98 (0.4, 10.5) 3.22 (0.4, 29.2) 0.44 (0.2, 1.2) 0.61 (0.2, 1.8) 6.89 (0.3, 400) 0.55 (0.2, 2) 0.47 (0.1, 1.6) 0.71 (0.1, 3.8) 0.5 (0.3, 0.8) 0.57 (0.17, 1.9) 0.28 (0.1, 0.7) 1.13 (0.2, 6.3) 0.23 (0.04, 1.3)
N=2, 1.11 (0.6, 2) N=1, 1.53 (0.6, 4.2)
M AN U
3.0 (1.9, 4.9) 4.26 (2.1, 8.8) 1.92 (1.1, 3.4) 2.66 (1.2, 6.2) 4.5 (2.2, 9.0) 1.56 (0.7, 3.7) 5.96 (1.1, 32.2) 9.72 (1.2, 89.2) 1.34 (0.5, 3.4) 1.82 (0.6, 5.5) 20.75 (0.9, 1213) 1.65 (0.5, 6.0) 1.41 (0.4, 5.1) 2.14 (0.4, 11.3) 1.50 (0.7, 3.0) 1.72 (0.5, 6.0) 0.85 (0.3, 2.3) 3.41 (0.6, 20.3) 0.68 (0.1, 4.1)
N=7, 2.7 (1.7, 4.3)
N=7, 1.68 (1.1, 2.5)
TE D
B
N=5, 2.56 (1.6, 4.1)
EP
0.82 (0.4, 1.6) 2.46 (1.3, 4.5) 3.48 (2.1,5.8) 1.57 (0.9, 2.8) 2.18 (1.3, 3.8) 3.64 (1.7, 7.8) 1.28 (0.6, 2.7) 4.88 (1.1, 23) 7.92 (1.1, 67) 1.09 (0.4, 3) 1.49 (0.6, 3.9) 16.83 (0.8, 947) 1.35 (0.4, 4.7) 1.15 (0.4, 3.4) 1.75 (0.3, 9.0) 1.23 (0.6, 2.6) 1.41 (0.5, 4.0) 0.7 (0.3, 1.9) 2.79 (0.4, 17.2) 0.56 (0.1, 2.9)
N=9, 2.58 (1.3, 5.3) N=3, 5.46 (0.8, 38.5)
AC C
A
RI PT
Table 2: Results (odds ratios, with 95% confidence intervals/credible intervals) of the network meta-analysis for global effect
L
11.41 (0.4, 699) 0.9 (0.2, 4.0) 0.77 (0.2, 3.3) 1.17 (0.2, 7.1) 0.82 (0.3, 2.6) 0.94 (0.2, 3.8) 0.47 (0.1, 1.8) 1.87 (0.2, 14.1) 0.37 (0.1, 2.5)
M 0.08 (0.0, 2.3) 0.07 (0.0, 1.8) 0.1 (0.0, 3.5) 0.07 (0.0, 1.7) 0.08 (0.0, 2.2) 0.04 (0.0, 1.1) 0.16 (0.0, 6.0) 0.03 (0.0, 1.1)
N=2, 1.31 (0.8, 2.3)
N 0.85 (0.2, 4.5) 1.30 (0.5, 3.7) 0.91 (0.2, 3.6) 1.04 (0.2, 5.4) 0.51 (0.1, 2.4) 2.07 (0.2, 17.52) 0.41 (0.05, 3.4)
N=2, 0.78 (0.4, 1.7)
O 1.52 (0.2, 10.8) 1.07 (0.3, 4) 1.23 (0.4, 3.6) 0.6 (0.1, 2.7) 2.42 (0.3, 20.3) 0.48 (0.1, 3.5)
P 0.70 (0.1, 3.9) 0.8 (0.1, 5.6) 0.4 (0.1, 2.6) 1.6 (0.1, 18) 0.32 (0.03, 3.3)
Q 1.15 (0.3, 4.2) 0.57 (0.2, 1.5) 2.27 (0.4, 13.8) 0.45 (0.1, 2.8)
R 0.50 (0.1, 2.1) 2.0 (0.2, 16.3) 0.40 (0.1, 2.9)
S 4.0 (0.6, 28) 0.80 (0.1, 5.7)
T 0.20 (0.02, 2.3)
Lower triangle includes the findings of the network meta-analysis (posterior median odds ratios ORs plus 95% credible intervals) conducted in the Bayesian statistical package WinBUGS; upper triangle includes the findings of the direct standard pairwise meta-analyses (OR plus confidence intervals) conducted using STATA Statistically significant findings have been shaded (significance assessment made on data rounded to decimal places)
1
U
ACCEPTED MANUSCRIPT
OR > 1.0 favours intervention compared with control. A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intra-operative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L
RI PT
Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; S intradiscal injections; T Spinal cord simulations; U
AC C
EP
TE D
M AN U
SC
Radiofrequency treatment; N number of studies included in conventional pairwise meta-analysis.
2
ACCEPTED MANUSCRIPT
Table 3: Results (weighted mean difference, with 95% confidence intervals/credible intervals) of the network meta-analysis for pain
-1.89 (-17, 14) 3.02 (-10, 16) 7.02 (-9, 23) -5.12 (-14, 4) 7.79 (-17, 33) 2.4 (-27, 32) -5.21 (-29, 18) 6.95 (-12, 26) 9.14 (-15, 34) -9.96 (-30, 10) 26.7 (-8, 62) 14.51 (-10, 40) 25.83 (-13, 65) 21.14 (-7, 49) -2.07 (-27, 23) 22.53 (-7, 52)
D 4.85 (-13, 23) 8.86 (-1, 19) -3.27 (-21, 15) 9.61 (-11, 31) 4.19 (-28, 36) -3.42 (-22, 15) 8.79 (-13, 31) 10.96 (-8, 31) -8.10 (-22, 6) 28.56 (-3, 61) 16.32 (-5, 38) 27.61 (-8, 64) 22.94 (-8, 53) -0.21 (-21, 21) 24.38 (-2, 51)
N=1, 1.60 (-8, 11) N=2,18.01 (6, 30)
N=8,-5.17 (-12, 2) N=1, 4.00 (-6, 14) N=1, -0.63 (-15, 14)
F -12.09 (-30, 6) 0.69 (-21, 23) -4.7 (-37, 27) -12.28 (-33, 8) -0.07 (-23, 23) 2.10 (-18, 23) -16.99 (-32, -3) 19.65 (-13, 53) 7.44 (-13, 27) 18.77 (-17, 56) 14.14 (-17, 45) -9.01 (-30, 12) 15.52 (-11, 42)
G 12.89 (-14, 40) 7.48 (-23, 38) -0.13 (-26, 25) 12.07 (-9, 33) 14.26 (-11, 40) -4.85 (-27, 17) 31.84 (-4, 68) 19.61 (-7, 46) 30.99 (-9, 71) 26.2 (-3, 55) 3.06 (-23, 30) 27.64 (-3, 58)
N=1, -7.00 (-14, -0.4)
N=2, -9.91 (-43, 23)
N=1, 35.00 (-25, 95)
N=1, -9.3 (-23, 4.9)
N=1, -2.00 (-12, 8) N=1, 9.00 (-4, 22)
E 4.04 (-15, 22) -8.11 (-24, 8) 4.79 (-22, 32) -0.63 (-27, 26) -8.25 (-34, 17) 3.95 (-19, 27) 6.15 (-19, 32) -12.95 (-35, 9) 23.65 (-12, 60) 11.47 (-15, 38) 22.8 (-17, 63) 18.14 (-6, 42) -5.07 (-31, 21) 19.52 (-11, 50)
N=1, -25.00 (-42, -8)
SC
N=3, 2.36 (-8, 13)
C
N= 1, 3.36 (-15, 21)
I -7.58 (-44, 29) 4.58 (-30, 39) 6.81 (-30, 44) -12.3 (-46, 22) 24.38 (-20, 69) 12.15 (-25, 49) 23.39 (-24, 71) 18.8 (-17, 55) -4.37 (-42, 33) 20.19 (-20, 61)
N=1, -26.66 (-38, -15)
N=1,13.00 (2, 24)
N=1, 18.00 (8, 28) N=1, -40.50 (-58, -23)
N=2, 3.19 (-13, 19)
H -5.42 (-43, 32) -13.02 (-41, 14) -0.75 (-31, 29) 1.4 (-14, 18) -17.72 (-42, 6) 19.02 (-6, 43) 6.74 (-22, 35) 18.11 (-12, 48) 13.38 (-23, 49) -9.82 (-38, 18) 14.84 (-17, 46)
M AN U
-7.13 (-21, 6) -8.96 (-23, 6) -4.17 (-22, 14) -0.10 (-16, 16) -12.21 (-29, 4) 0.68 (-24, 26) -4.77 (-37, 27) -12.35 (-36, 11) -0.19 (-19, 19) 1.98 (-21, 26) -17.03 (-37, 2) 19.62 (-15, 55) 7.41 (-18, 32) 18.71 (-20, 57) 13.97 (-16, 44) -9.13 (-34, 16) 15.36 (-14, 45)
N=2, -7.0 (-12, -2)
N=4, -6.26 (-15, 3)
N=1, 22.50 (10, 35)
N=1, 19.00 (8, 30)
J
12.19 (-16, 41) 14.45 (-12, 41) -4.71 (-27, 18) 32.01 (-4, 69) 19.75 (-8, 47) 31.07 (-9, 71) 26.36 (-9, 62) 3.21 (-24, 31) 27.76 (-3, 59)
TE D
B
N=1, -5.40 (-24, 13)
EP
-2.44 (-18, 13) -9.54 (-25, 6) -11.4 (-19, -4) -6.62 (-25, 11) -2.54 (-12, 6) -14.64 (-33, 3) -1.81 (-22, 18) -7.24 (-39, 25) -14.82 (-34, 4) -2.62 (-25, 20) -0.46 (-18, 18) -19.51 (-33, -6) 17.13 (-14, 49) 4.92 (-16, 25) 16.22 (-19, 52) 11.6 (-19, 42) -11.6 (-32, 8) 12.98 (-12, 38)
N=7, -8.11 (-19, 3) N=2, -5.32 (-12, 1)
AC C
A
RI PT
intensity
K
2.16 (-26, 31) -16.93 (-43, 8) 19.77 (-18, 58) 7.58 (-22, 37) 18.9 (-23, 61) 14.17 (-19, 47) -8.99 (-39, 21) 15.55 (-18, 49)
L
-19.09 (-42, 3) 17.62 (-12, 47) 5.42 (-23, 32) 16.69 (-17, 50) 11.96 (-24, 47) -11.13 (-39, 16) 13.42 (-18, 44)
M 36.68 (3, 71) 24.45 (0.78, 48) 35.76 (-2, 74) 31.04 (-2, 64) 7.93 (-16, 32) 32.48 (5, 61)
N=2, -1.09 (-7, 5)
N -12.19 (-50, 25) -0.90 (-18, 16) -5.46 (-49, 37) -28.78 (-66, 8) -4.16 (-44, 35)
N=2, -8.23 (-18, 2)
O 11.27 (-29, 53) 6.55 (-29 43) -16.52 (-33, 0.26) 8.01 (-24, 40)
P -4.62 (-52, 41) -27.81 (-69, 12) -3.28 (-46, 40)
Q -23.11 (-59 13) 1.4 (-38, 41)
R 24.55 (-7, 57)
Lower triangle includes the findings of the network meta-analysis (posterior median weighted mean differences WMDs plus 95% credible intervals) conducted in the Bayesian statistical package WinBUGS; upper triangle includes the findings of the direct standard pairwise meta-analyses (WMD plus confidence intervals) conducted using STATA Statistically significant findings have been shaded (significance assessment made on data rounded to decimal places) WMD > 0 (representing reduction in pain) favours intervention compared with control
1
U
ACCEPTED MANUSCRIPT
A Inactive control; B Conventional care; C Disc surgery; D Epidural injections; E Chemonucleolysis; F Non-opioids; G Intra-operative interventions; H Traction; I Manipulation; J Acupuncture; K Exercise therapy; L Passive physical therapy; M Biological agents; N Bed rest; O Opioids; P Education/Advice; Q Percutaneous discectomy; R Neuropathic pain modulators; U Radiofrequency treatment; N number of studies
AC C
EP
TE D
M AN U
SC
RI PT
included in conventional pairwise meta-analysis.
2
ACCEPTED MANUSCRIPT
Figure 1: flow diagram showing the number of references identified, publications retrieved for assessment, and studies included in the review
References identified by electronic searches after de-duplication N=33560 References obtained from other sources (searching bibliographies, reference lists of reviews) N=30
RI PT
References rejected on the basis of title and/or abstract N=32812
Ongoing/completed studies with no available outcome data N=42 (Most identified through trial registries)
AC C
EP
TE D
Studies included in the HTA review* N=270 (Publications N=388)
Studies that reported data on back specific functional status but did not report useable data on global effect or pain intensity N=5
Studies included in analysis of global effect N=95 (68 RCTs/Q-RCTs)
Publications excluded from HTA review* because they did not meet criteria relating to design, population, intervention or outcome N=241
M AN U
Publications that could not be assessed for inclusion because: Unable to translate N=93 Unavailable from interlibrary loans N=14
SC
Publications retrieved in full for detailed evaluation and assessment for inclusion N=777
Studies included in refined network meta-analyses N=122 (90 RCTs/Q-RCTs)
Studies excluded from initial network metaanalyses because they evaluated mixed treatments, postoperative interventions or interventions classified under the same treatment strategy N=142 Studies excluded from refined network metaanalyses: experimental intervention no longer part of the new nonopioid category N=3
Studies included in analysis of pain intensity N=53 (46 RCTs/Q-RCTs)
*The current study represents further refinements of initial network meta-analyses conducted as part of a broader Health Technology Assessment (HTA) review evaluating the clinical and cost effectiveness of sciatica treatments;16 the original review included a narrative synthesis, conventional pairwise analyses, network meta-analyses, a review of economic evaluations and an economic evaluation. Five studies reported data on back specific functional status (an outcome considered as part of the initial network meta-analyses), but did not report useable data for global effect or pain intensity.
ACCEPTED MANUSCRIPT
Figure 2: Network of treatment strategies for sciatica for comparative studies reporting global effect
RI PT
Each treatment strategy is a node in the network. The links between the nodes are arms in head-to-head (direct) comparisons in eligible studies. The numbers along the link lines indicate the number of studies or pairs of study arms for that link in the network, with the observed I2 statistic (based on the data from the pairwise meta-analyses) is presented in brackets (0-40% might not be important; 30-60% moderate heterogeneity; 50-90% substantial heterogeneity 75% to 100% considerable heterogeneity34). Links that do not have any randomised controlled trial (RCT) or Quasi-
AC C
EP
TE D
M AN U
SC
RCT evidence are indicated in green.
Biological agents (M)
Manipulation (I) 1
Radiofrequency treatment (U)
Inactive control (A)
1
1
1
9 (70%)
1
Epidural injections / Nerve block (D) 3 (77%) 1
TE D
1 2 (0%)
Passive physical therapy (L)
1
1
Exercise therapy (K)
Education / Advice (P)
2 (0%)
2 (0%) Opioids (O)
Conventional care (B) 7 (79%) Disc surgery (C) 7 (0%)
1
1 5 (20%)
AC C
1
EP
1 Traction (H)
SC 1
M AN U
Acupuncture (J)
Neuropathic painmodulators (R)
2 (38%)
7 (23%)
Non-opioids (F)
1
1
RI PT
ACCEPTED MANUSCRIPT
14 (68%)
5 (74%) Percutaneous discectomy (Q)
1
Spinal cord stimulation (T)
5 (49%)
1
1
Bed rest (N)
4 (77%) Intra-operative interventions (G)
Chemonucleolysis (E) 4 (57%) Intradiscal injections (S)
ACCEPTED MANUSCRIPT
Figure 3: Network of treatment strategies for sciatica for comparative studies reporting pain intensity
RI PT
Each treatment strategy is a node in the network. The links between the nodes are arms in head-to-head (direct) comparisons in eligible studies. The numbers along the link lines indicate the number of studies or pairs of study arms for that link in the network, with the observed I2 statistic (based on the data from the pairwise meta-analyses) is presented in brackets (0-40% might not be important; 30-60% moderate heterogeneity; 50-90% substantial heterogeneity 75% to 100% considerable heterogeneity34). Links that do not have any randomised controlled trial (RCT) or Quasi-
AC C
EP
TE D
M AN U
SC
RCT evidence are indicated in green.
ACCEPTED MANUSCRIPT
RI PT
Radiofrequency treatment (U)
1
Inactive control (A)
Opioids (O)
8 (77%)
Conventional care (B)
1
1
Non-opioids (F)
2 (0%)
Epidural injections / Nerve block (D)
1
EP
1
TE D
7 (89%)
2 (87%)
Passive Physical Therapy (L)
2 (2%) 1 Disc surgery (C)
1
Percutaneous discectomy (Q)
3 (68%)
1
1
Chemonucleolysis (E)
1
1 1
AC C
1
1
2 (74%)
Traction (H) Education / Advice (P)
Biological agents (M)
1
4 (84%)
Intra-operative interventions (G)
SC
2 (0%)
1
M AN U
Neuropathic painmodulators (R)
2 (93%)
2 (0%)
Acupuncture (J)
1 Bed rest (N)
Exercise therapy (K)
Manipulation (I)
ACCEPTED MANUSCRIPT
Figure 4: Plot of the odds ratios (ORs) of global effects for different treatment strategies compared with inactive control from the network meta-analysis NB The data has been spun around so that effect estimates that favour the intervention are shown on the right hand side. This means that the an OR 0 represents a reduction in pain intensity in favour of the intervention
Treatment
ES (95% CI)
Biological agents (M)
-19.58 (-32.69, -6.47)
Acupuncture (J)
-14.87 (-33.60, 3.86)
SC
category
Intra-operative interventions (G)
-14.71 (-32.66, 3.24) -11.62 (-31.49, 8.25)
Neuropathic painmodulators (R)
-11.43 (-19.12, -3.74)
Epidural injections (D)
M AN U
Disc surgery (C) Manipulation (I) Chemonucleolysis (E) Exercise therapy (K) Non-opioids (F) Conventional care (B) Traction (H) Passive physical therapy (L) Opioids (O) Percutaneous discectomy (Q)
Education/Advice (P) Bed rest (N)
TE D
Radiofrequency treatment (U)
-9.60 (-25.25, 6.04) -7.22 (-38.90, 24.45) -6.59 (-24.35, 11.17) -2.63 (-24.97, 19.72) -2.58 (-11.61, 6.44) -2.45 (-17.78, 12.88) -1.77 (-21.55, 18.00) -0.35 (-18.50, 17.80) 4.89 (-15.43, 25.22) 11.50 (-18.57, 41.57) 12.96 (-11.74, 37.66) 16.32 (-18.76, 51.40) 17.22 (-13.77, 48.21)
NOTE: Weights are from random effects analysis
-40
-30
-20
-10 -5 0 5 10
20
30
40
favours inactive control
AC C
EP
favours intervention
1
